Download Full Version ebook - Visit ebooknice.

com

(Ebook) Python deeper insights into machine

learning: leverage benefits of machine learning
techniques using Python: a course in three modules
by Hearty, John;Julian, David;Raschka, Sebastian
ISBN 9781787128545, 9781787128576, 1787128547,
https://ebooknice.com/product/python-deeper-insights-into-
1787128571
machine-learning-leverage-benefits-of-machine-learning-
techniques-using-python-a-course-in-three-modules-11860614

Click the button below to download

DOWLOAD EBOOK

Discover More Ebook - Explore Now at ebooknice.com

 Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Start reading on any device today!

(Ebook) Python: Deeper Insights into Machine Learning:

Leverage benefits of machine learning techniques using
Python by Raschka, Sebastian, Julian, David, Hearty, John
ISBN 9781787128545, 1787128547
https://ebooknice.com/product/python-deeper-insights-into-machine-
learning-leverage-benefits-of-machine-learning-techniques-using-
python-29785114
ebooknice.com

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason;

Viles, James ISBN 9781459699816, 9781743365571,
9781925268492, 1459699815, 1743365578, 1925268497
https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

ebooknice.com

(Ebook) Matematik 5000+ Kurs 2c Lärobok by Lena

Alfredsson, Hans Heikne, Sanna Bodemyr ISBN 9789127456600,
9127456609
https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312

ebooknice.com

(Ebook) SAT II Success MATH 1C and 2C 2002 (Peterson's SAT

II Success) by Peterson's ISBN 9780768906677, 0768906679

https://ebooknice.com/product/sat-ii-success-
math-1c-and-2c-2002-peterson-s-sat-ii-success-1722018

ebooknice.com
(Ebook) Master SAT II Math 1c and 2c 4th ed (Arco Master
the SAT Subject Test: Math Levels 1 & 2) by Arco ISBN
9780768923049, 0768923042
https://ebooknice.com/product/master-sat-ii-math-1c-and-2c-4th-ed-
arco-master-the-sat-subject-test-math-levels-1-2-2326094

ebooknice.com

(Ebook) Cambridge IGCSE and O Level History Workbook 2C -

Depth Study: the United States, 1919-41 2nd Edition by
Benjamin Harrison ISBN 9781398375147, 9781398375048,
1398375144, 1398375047
https://ebooknice.com/product/cambridge-igcse-and-o-level-history-
workbook-2c-depth-study-the-united-states-1919-41-2nd-edition-53538044

ebooknice.com

(Ebook) Python Real World Machine Learning: Real World

Machine Learning: Take your Python Machine learning skills
to the next level by Joshi, Prateek, Massaron, Luca,
Hearty, John ISBN 9781787123212, 1787123219
https://ebooknice.com/product/python-real-world-machine-learning-real-
world-machine-learning-take-your-python-machine-learning-skills-to-
the-next-level-5602660
ebooknice.com

(Ebook) Python Machine Learning. Machine Learning and Deep

Learning with Python, scikit-learn and TensorFlow by
Sebastian Raschka, Vahid Mirjalili ISBN 9781787125933,
1787125939
https://ebooknice.com/product/python-machine-learning-machine-
learning-and-deep-learning-with-python-scikit-learn-and-
tensorflow-6868034
ebooknice.com

(Ebook) Transactional Machine Learning with Data Streams

and AutoML: Build Frictionless and Elastic Machine
Learning Solutions with Apache Kafka in the Cloud Using
Python by Sebastian Maurice ISBN 9781484270226, 1484270223
https://ebooknice.com/product/transactional-machine-learning-with-
data-streams-and-automl-build-frictionless-and-elastic-machine-
learning-solutions-with-apache-kafka-in-the-cloud-using-
python-37321800
ebooknice.com
Python: Deeper Insights
into Machine Learning
Leverage benefits of machine learning techniques
using Python

</>

LEARNING PATH
Python: Deeper Insights into
Machine Learning

Leverage benefits of machine learning

techniques using Python

A course in three modules

BIRMINGHAM - MUMBAI
Python: Deeper Insights into Machine Learning

Copyright © 2016 Packt Publishing

All rights reserved. No part of this course may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, without the prior written
permission of the publisher, except in the case of brief quotations embedded in
critical articles or reviews.

Every effort has been made in the preparation of this course to ensure the accuracy
of the information presented. However, the information contained in this course
is sold without warranty, either express or implied. Neither the authors, nor Packt
Publishing, and its dealers and distributors will be held liable for any damages
caused or alleged to be caused directly or indirectly by this course.

Packt Publishing has endeavored to provide trademark information about all of the
companies and products mentioned in this course by the appropriate use of capitals.
However, Packt Publishing cannot guarantee the accuracy of this information.

Published on: August 2016

Published by Packt Publishing Ltd.

Livery Place
35 Livery Street
Birmingham B3 2PB, UK.

ISBN 978-1-78712-857-6

www.packtpub.com
 Credits

Authors Content Development Editor

Sebastian Raschka Amrita Noronha
David Julian
John Hearty Production Coordinator
Arvindkumar Gupta

Reviewers
Richard Dutton
Dave Julian
Vahid Mirjalili
Hamidreza Sattari
Dmytro Taranovsky
Dr. Vahid Mirjalili
Jared Huffman
Ashwin Pajankar
 Preface
Machine learning and predictive analytics are becoming one of the key strategies for
unlocking growth in a challenging contemporary marketplace .It is one of the fastest
growing trends in modern computing and everyone wants to get into the field of
machine learning. In order to obtain sufficient recognition in this field, one must be
able to understand and design a machine learning system that serves the needs of a
project. The idea is to prepare a Learning Path that will help you to tackle the real-
world complexities of modern machine learning with innovative and cutting-edge
techniques. Also, it will give you a solid foundation in the machine learning design
process, and enable you to build customized machine learning models to solve
unique problems

What this learning path covers

Module 1, Python Machine Learning, discusses the essential machine algorithms for
classification and provides practical examples using scikit-learn. It teaches you to
prepare variables of different types and also speaks about polynomial regression
and tree-based approaches. This module focuses on open source Python library that
allows us to utilize multiple cores of modern GPUs.

Module 2, Designing Machine Learning Systems with Python, acquaints you with large
library of packages for machine learning tasks. It introduces broad topics such
as big data, data properties, data sources, and data processing .You will further
explore models that form the foundation of many advanced nonlinear techniques.
This module will help you in understanding model selection and parameter tuning
techniques that could help in various case studies.

[i]
Preface

Module 3, Advanced Machine Learning with Python, helps you to build your skill
with deep architectures by using stacked denoising autoencoders. This module is a
blend of semi-supervised learning techniques, RBM and DBN algorithms .Further
this focuses on tools and techniques which will help in making consistent working
process.

What you need for this learning path

Module 1, Python Machine Learning will require an installation of Python 3.4.3 or
newer on Mac OS X, Linux or Microsoft Windows. Use of Python essential libraries
like SciPy, NumPy, scikit-Learn, matplotlib, and pandas. is essential.

Before you start, Please refer:

• The direct link to the Iris dataset would be: https://raw.

githubusercontent.com/rasbt/python-machine-learning-book/
master/code/datasets/iris/iris.data
• We've added some additional notes to the code notebooks mentioning the
offline datasets in case there are server errors. https://www.dropbox.com/
sh/tq2qdh0oqfgsktq/AADIt7esnbiWLOQODn5q_7Dta?dl=0
• Module 2, Designing Machine Learning Systems with Python, will need an
inclination to learn machine learning and the Python V3 software, which you
can download from https://www.python.org/downloads/.
• Module 3, Advanced Machine Learning with Python, leverages openly
available data and code, including open source Python libraries and
frameworks.

Who this learning path is for

This title is for Data scientist and researchers who are already into the field of
Data Science and want to see Machine learning in action and explore its real-world
application. Prior knowledge of Python programming and mathematics is must with
basic knowledge of machine learning concepts.

Reader feedback
Feedback from our readers is always welcome. Let us know what you think about
this course—what you liked or disliked. Reader feedback is important for us as it
helps us develop titles that you will really get the most out of.

[ ii ]
 Preface

To send us general feedback, simply e-mail feedback@packtpub.com, and mention

the course's title in the subject of your message.

If there is a topic that you have expertise in and you are interested in either writing
or contributing to a book, see our author guide at www.packtpub.com/authors.

Customer support
Now that you are the proud owner of a Packt course, we have a number of things to
help you to get the most from your purchase.

Downloading the example code

You can download the example code files for this course from your account at
http://www.packtpub.com. If you purchased this course elsewhere, you can visit
http://www.packtpub.com/support and register to have the files e-mailed directly
to you.

You can download the code files by following these steps:

1. Log in or register to our website using your e-mail address and password.
2. Hover the mouse pointer on the SUPPORT tab at the top.
3. Click on Code Downloads & Errata.
4. Enter the name of the course in the Search box.
5. Select the course for which you're looking to download the code files.
6. Choose from the drop-down menu where you purchased this course from.
7. Click on Code Download.

You can also download the code files by clicking on the Code Files button on the
course's webpage at the Packt Publishing website. This page can be accessed by
entering the course's name in the Search box. Please note that you need to be logged
in to your Packt account.

Once the file is downloaded, please make sure that you unzip or extract the folder
using the latest version of:

• WinRAR / 7-Zip for Windows

• Zipeg / iZip / UnRarX for Mac
• 7-Zip / PeaZip for Linux

[ iii ]
Preface

The code bundle for the course is also hosted on GitHub at https://github.com/
PacktPublishing/Python-Deeper-Insights-into-Machine-Learning.

Errata
Although we have taken every care to ensure the accuracy of our content, mistakes
do happen. If you find a mistake in one of our courses—maybe a mistake in the text
or the code—we would be grateful if you could report this to us. By doing so, you
can save other readers from frustration and help us improve subsequent versions
of this course. If you find any errata, please report them by visiting http://www.
packtpub.com/submit-errata, selecting your course, clicking on the Errata
Submission Form link, and entering the details of your errata. Once your errata are
verified, your submission will be accepted and the errata will be uploaded to our
website or added to any list of existing errata under the Errata section of that title.

To view the previously submitted errata, go to https://www.packtpub.com/books/

content/support and enter the name of the course in the search field. The required
information will appear under the Errata section.

Piracy
Piracy of copyrighted material on the Internet is an ongoing problem across all
media. At Packt, we take the protection of our copyright and licenses very seriously.
If you come across any illegal copies of our works in any form on the Internet, please
provide us with the location address or website name immediately so that we can
pursue a remedy.

Please contact us at copyright@packtpub.com with a link to the suspected

pirated material.

We appreciate your help in protecting our authors and our ability to bring you
valuable content.

Questions
If you have a problem with any aspect of this course, you can contact us at
questions@packtpub.com, and we will do our best to address the problem.

[ iv ]
Course Module 1: Python Machine Learning

Chapter 1: Giving Computers the Ability to Learn from Data 3

Building intelligent machines to transform data into knowledge 4
The three different types of machine learning 4
An introduction to the basic terminology and notations 10
A roadmap for building machine learning systems 12
Using Python for machine learning 15
Summary 17
Chapter 2: Training Machine Learning Algorithms
for Classification 19
Artificial neurons – a brief glimpse into the early history of
machine learning 20
Implementing a perceptron learning algorithm in Python 26
Adaptive linear neurons and the convergence of learning 35
Summary 49
Chapter 3: A Tour of Machine Learning Classifiers
Using Scikit-learn 51
Choosing a classification algorithm 51
First steps with scikit-learn 52
Modeling class probabilities via logistic regression 58
Maximum margin classification with support vector machines 71
Solving nonlinear problems using a kernel SVM 77
Decision tree learning 82

[i]
Table of Contents

K-nearest neighbors – a lazy learning algorithm 94

Summary 98
Chapter 4: Building Good Training Sets – Data Preprocessing 101
Dealing with missing data 101
Handling categorical data 106
Partitioning a dataset in training and test sets 110
Bringing features onto the same scale 112
Selecting meaningful features 114
Assessing feature importance with random forests 126
Summary 128
Chapter 5: Compressing Data via Dimensionality Reduction 129
Unsupervised dimensionality reduction via principal
component analysis 130
Supervised data compression via linear discriminant analysis 140
Using kernel principal component analysis for nonlinear mappings 150
Summary 169
Chapter 6: Learning Best Practices for Model Evaluation
and Hyperparameter Tuning 171
Streamlining workflows with pipelines 171
Using k-fold cross-validation to assess model performance 175
Debugging algorithms with learning and validation curves 181
Fine-tuning machine learning models via grid search 187
Looking at different performance evaluation metrics 191
Summary 200
Chapter 7: Combining Different Models for Ensemble Learning 201
Learning with ensembles 201
Implementing a simple majority vote classifier 205
Evaluating and tuning the ensemble classifier 215
Bagging – building an ensemble of classifiers from
bootstrap samples 221
Leveraging weak learners via adaptive boosting 226
Summary 234
Chapter 8: Applying Machine Learning to Sentiment Analysis 235
Obtaining the IMDb movie review dataset 235
Introducing the bag-of-words model 238
Training a logistic regression model for document classification 246
Working with bigger data – online algorithms and out-of-core learning 248
Summary 252

[ ii ]
 Table of Contents

Chapter 9: Embedding a Machine Learning Model into a

Web Application 253
Serializing fitted scikit-learn estimators 254
Setting up a SQLite database for data storage 257
Developing a web application with Flask 259
Turning the movie classifier into a web application 266
Deploying the web application to a public server 274
Summary 278
Chapter 10: Predicting Continuous Target Variables with
Regression Analysis 279
Introducing a simple linear regression model 280
Exploring the Housing Dataset 281
Implementing an ordinary least squares linear regression model 287
Fitting a robust regression model using RANSAC 293
Evaluating the performance of linear regression models 296
Using regularized methods for regression 299
Turning a linear regression model into a curve – polynomial
regression 300
Summary 311
Chapter 11: Working with Unlabeled Data – Clustering Analysis 313
Grouping objects by similarity using k-means 314
Organizing clusters as a hierarchical tree 328
Locating regions of high density via DBSCAN 336
Summary 342
Chapter 12: Training Artificial Neural Networks for
Image Recognition 343
Modeling complex functions with artificial neural networks 344
Classifying handwritten digits 352
Training an artificial neural network 367
Developing your intuition for backpropagation 374
Debugging neural networks with gradient checking 375
Convergence in neural networks 381
Other neural network architectures 383
A few last words about neural network implementation 386
Summary 387

[ iii ]
Table of Contents

Chapter 13: Parallelizing Neural Network Training with Theano 389

Building, compiling, and running expressions with Theano 390
Choosing activation functions for feedforward neural networks 403
Training neural networks efficiently using Keras 410
Summary 416

Course Module 2: Designing Machine Learning

Systems with Python

Chapter 1: Thinking in Machine Learning 421

The human interface 422
Design principles 425
Summary 453
Chapter 2: Tools and Techniques 455
Python for machine learning 456
IPython console 456
Installing the SciPy stack 457
NumPY 458
Matplotlib 464
Pandas 468
SciPy 471
Scikit-learn 474
Summary 481
Chapter 3: Turning Data into Information 483
What is data? 484
Big data 484
Signals 500
Cleaning data 502
Visualizing data 504
Summary 507
Chapter 4: Models – Learning from Information 509
Logical models 509
Tree models 517
Rule models 521
Summary 528

[ iv ]
 Table of Contents

Chapter 5: Linear Models 529

Introducing least squares 530
Logistic regression 538
Multiclass classification 544
Regularization 545
Summary 548
Chapter 6: Neural Networks 549
Getting started with neural networks 549
Logistic units 551
Cost function 556
Implementing a neural network 559
Gradient checking 565
Other neural net architectures 566
Summary 567
Chapter 7: Features – How Algorithms See the World 569
Feature types 570
Operations and statistics 571
Structured features 574
Transforming features 574
Principle component analysis 583
Summary 585
Chapter 8: Learning with Ensembles 587
Ensemble types 587
Bagging 588
Boosting 594
Ensemble strategies 601
Summary 604
Chapter 9: Design Strategies and Case Studies 605
Evaluating model performance 605
Model selection 610
Learning curves 613
Real-world case studies 615
Machine learning at a glance 626
Summary 627

[v]
Table of Contents

Course Module 3: Advanced Machine

Learning with Python

Chapter 1: Unsupervised Machine Learning 631

Principal component analysis 632
Introducing k-means clustering 637
Self-organizing maps 648
Further reading 654
Summary 655
Chapter 2: Deep Belief Networks 657
Neural networks – a primer 658
Restricted Boltzmann Machine 663
Deep belief networks 679
Further reading 685
Summary 686
Chapter 3: Stacked Denoising Autoencoders 687
Autoencoders 687
Stacked Denoising Autoencoders 696
Further reading 705
Summary 705
Chapter 4: Convolutional Neural Networks 707
Introducing the CNN 707
Further Reading 729
Summary 730
Chapter 5: Semi-Supervised Learning 731
Introduction 731
Understanding semi-supervised learning 732
Semi-supervised algorithms in action 733
Further reading 756
Summary 757
Chapter 6: Text Feature Engineering 759
Introduction 759
Text feature engineering 760
Further reading 783
Summary 784
Chapter 7: Feature Engineering Part II 785
Introduction 785
Creating a feature set 786

[ vi ]
 Table of Contents

Feature engineering in practice 805

Further reading 829
Summary 830
Chapter 8: Ensemble Methods 831
Introducing ensembles 832
Using models in dynamic applications 851
Further reading 863
Summary 864
Chapter 9: Additional Python Machine Learning Tools 865
Alternative development tools 866
Further reading 875
Summary 875
Chapter 10: Chapter Code Requirements 879
Biblography 881

[ vii ]
 Module 1

Python Machine Learning

Leverage benefits of machine learning techniques using Python

 Giving Computers the Ability
to Learn from Data
In my opinion, machine learning, the application and science of algorithms that makes
sense of data, is the most exciting field of all the computer sciences! We are living in
an age where data comes in abundance; using the self-learning algorithms from the
field of machine learning, we can turn this data into knowledge. Thanks to the many
powerful open source libraries that have been developed in recent years, there has
probably never been a better time to break into the machine learning field and learn
how to utilize powerful algorithms to spot patterns in data and make predictions
about future events.

In this chapter, we will learn about the main concepts and different types of machine
learning. Together with a basic introduction to the relevant terminology, we will lay
the groundwork for successfully using machine learning techniques for practical
problem solving.

In this chapter, we will cover the following topics:

• The general concepts of machine learning

• The three types of learning and basic terminology
• The building blocks for successfully designing machine learning systems
• Installing and setting up Python for data analysis and machine learning

[3]
Giving Computers the Ability to Learn from Data

Building intelligent machines to

transform data into knowledge
In this age of modern technology, there is one resource that we have in abundance: a
large amount of structured and unstructured data. In the second half of the twentieth
century, machine learning evolved as a subfield of artificial intelligence that involved
the development of self-learning algorithms to gain knowledge from that data in
order to make predictions. Instead of requiring humans to manually derive rules
and build models from analyzing large amounts of data, machine learning offers a
more efficient alternative for capturing the knowledge in data to gradually improve
the performance of predictive models, and make data-driven decisions. Not only is
machine learning becoming increasingly important in computer science research but
it also plays an ever greater role in our everyday life. Thanks to machine learning,
we enjoy robust e-mail spam filters, convenient text and voice recognition software,
reliable Web search engines, challenging chess players, and, hopefully soon, safe and
efficient self-driving cars.

The three different types of

machine learning
In this section, we will take a look at the three types of machine learning: supervised
learning, unsupervised learning, and reinforcement learning. We will learn about the
fundamental differences between the three different learning types and, using
conceptual examples, we will develop an intuition for the practical problem
domains where these can be applied:

[4]
 Chapter 1

Making predictions about the future with

supervised learning
The main goal in supervised learning is to learn a model from labeled training data
that allows us to make predictions about unseen or future data. Here, the term
supervised refers to a set of samples where the desired output signals (labels) are
already known.

Considering the example of e-mail spam filtering, we can train a model using a
supervised machine learning algorithm on a corpus of labeled e-mail, e-mail that are
correctly marked as spam or not-spam, to predict whether a new e-mail belongs to
either of the two categories. A supervised learning task with discrete class labels, such
as in the previous e-mail spam-filtering example, is also called a classification task.
Another subcategory of supervised learning is regression, where the outcome signal is
a continuous value:

Classification for predicting class labels

Classification is a subcategory of supervised learning where the goal is to
predict the categorical class labels of new instances based on past observations.
Those class labels are discrete, unordered values that can be understood as the
group memberships of the instances. The previously mentioned example of
e-mail-spam detection represents a typical example of a binary classification
task, where the machine learning algorithm learns a set of rules in order to
distinguish between two possible classes: spam and non-spam e-mail.

[5]
Giving Computers the Ability to Learn from Data

However, the set of class labels does not have to be of a binary nature. The predictive
model learned by a supervised learning algorithm can assign any class label that was
presented in the training dataset to a new, unlabeled instance. A typical example of
a multi-class classification task is handwritten character recognition. Here, we could
collect a training dataset that consists of multiple handwritten examples of each letter
in the alphabet. Now, if a user provides a new handwritten character via an input
device, our predictive model will be able to predict the correct letter in the alphabet
with certain accuracy. However, our machine learning system would be unable to
correctly recognize any of the digits zero to nine, for example, if they were not part
of our training dataset.

The following figure illustrates the concept of a binary classification task given 30
training samples: 15 training samples are labeled as negative class (circles) and 15
training samples are labeled as positive class (plus signs). In this scenario, our dataset
is two-dimensional, which means that each sample has two values associated with
it: x1 and x2 . Now, we can use a supervised machine learning algorithm to learn a
rule—the decision boundary represented as a black dashed line—that can separate
those two classes and classify new data into each of those two categories given its x1
and x2 values:

Regression for predicting continuous outcomes

We learned in the previous section that the task of classification is to assign
categorical, unordered labels to instances. A second type of supervised learning is
the prediction of continuous outcomes, which is also called regression analysis. In
regression analysis, we are given a number of predictor (explanatory) variables and a
continuous response variable (outcome), and we try to find a relationship between
those variables that allows us to predict an outcome.

[6]
 Chapter 1

For example, let's assume that we are interested in predicting the Math SAT
scores of our students. If there is a relationship between the time spent studying
for the test and the final scores, we could use it as training data to learn a model
that uses the study time to predict the test scores of future students who are
planning to take this test.

The term regression was devised by Francis Galton in his article

Regression Towards Mediocrity in Hereditary Stature in 1886. Galton
described the biological phenomenon that the variance of height
in a population does not increase over time. He observed that
the height of parents is not passed on to their children but the
children's height is regressing towards the population mean.

The following figure illustrates the concept of linear regression. Given a predictor
variable x and a response variable y, we fit a straight line to this data that minimizes
the distance—most commonly the average squared distance—between the sample
points and the fitted line. We can now use the intercept and slope learned from this
data to predict the outcome variable of new data:

[7]
Giving Computers the Ability to Learn from Data

Solving interactive problems with

reinforcement learning
Another type of machine learning is reinforcement learning. In reinforcement
learning, the goal is to develop a system (agent) that improves its performance
based on interactions with the environment. Since the information about the current
state of the environment typically also includes a so-called reward signal, we can
think of reinforcement learning as a field related to supervised learning. However, in
reinforcement learning this feedback is not the correct ground truth label or value,
but a measure of how well the action was measured by a reward function. Through
the interaction with the environment, an agent can then use reinforcement learning
to learn a series of actions that maximizes this reward via an exploratory
trial-and-error approach or deliberative planning.

A popular example of reinforcement learning is a chess engine. Here, the agent decides
upon a series of moves depending on the state of the board (the environment), and the
reward can be defined as win or lose at the end of the game:

Discovering hidden structures with

unsupervised learning
In supervised learning, we know the right answer beforehand when we train
our model, and in reinforcement learning, we define a measure of reward for
particular actions by the agent. In unsupervised learning, however, we are dealing
with unlabeled data or data of unknown structure. Using unsupervised learning
techniques, we are able to explore the structure of our data to extract meaningful
information without the guidance of a known outcome variable or reward function.

[8]
 Chapter 1

Finding subgroups with clustering

Clustering is an exploratory data analysis technique that allows us to organize a
pile of information into meaningful subgroups (clusters) without having any prior
knowledge of their group memberships. Each cluster that may arise during the
analysis defines a group of objects that share a certain degree of similarity but are
more dissimilar to objects in other clusters, which is why clustering is also sometimes
called "unsupervised classification." Clustering is a great technique for structuring
information and deriving meaningful relationships among data, For example, it
allows marketers to discover customer groups based on their interests in order to
develop distinct marketing programs.

The figure below illustrates how clustering can be applied to organizing unlabeled
data into three distinct groups based on the similarity of their features x1 and x2 :

Dimensionality reduction for data compression

Another subfield of unsupervised learning is dimensionality reduction. Often we are
working with data of high dimensionality—each observation comes with a high
number of measurements—that can present a challenge for limited storage space
and the computational performance of machine learning algorithms. Unsupervised
dimensionality reduction is a commonly used approach in feature preprocessing
to remove noise from data, which can also degrade the predictive performance of
certain algorithms, and compress the data onto a smaller dimensional subspace
while retaining most of the relevant information.

[9]
Giving Computers the Ability to Learn from Data

Sometimes, dimensionality reduction can also be useful for visualizing data—for

example, a high-dimensional feature set can be projected onto one-, two-, or
three-dimensional feature spaces in order to visualize it via 3D- or 2D-scatterplots
or histograms. The figure below shows an example where non-linear dimensionality
reduction was applied to compress a 3D Swiss Roll onto a
new 2D feature subspace:

An introduction to the basic terminology

and notations
Now that we have discussed the three broad categories of machine
learning—supervised, unsupervised, and reinforcement learning—let us have
a look at the basic terminology that we will be using in the next chapters. The
following table depicts an excerpt of the Iris dataset, which is a classic example in
the field of machine learning. The Iris dataset contains the measurements of 150 iris
flowers from three different species: Setosa, Versicolor, and Virginica. Please check if
this is replaced. Here, each flower sample represents one row in our data set, and the
flower measurements in centimeters are stored as columns, which we also call the
features of the dataset:

[ 10 ]
 Chapter 1

To keep the notation and implementation simple yet efficient, we will make use of
some of the basics of linear algebra. In the following chapters, we will use a matrix
and vector notation to refer to our data. We will follow the common convention to
represent each sample as separate row in a feature matrix X , where each feature is
stored as a separate column.

The Iris dataset, consisting of 150 samples and 4 features, can then be written as a
150 × 4 matrix X ∈ 150×4 :

[ 11 ]
Giving Computers the Ability to Learn from Data

For the rest of this book, we will use the superscript (i) to refer to the ith
training sample, and the subscript j to refer to the jth dimension of the
training dataset.
We use lower-case, bold-face letters to refer to vectors ( x ∈ R ) and
n×1

upper-case, bold-face letters to refer to matrices, respectively ( X ∈  n×m ) .

To refer to single elements in a vector or matrix, we write the letters in
(n) (n)
italics ( x or x( m ) , respectively).
For example, x1150 refers to the first dimension of flower sample 150, the
sepal length. Thus, each row in this feature matrix represents one flower
(i ) 1×4
instance and can be written as four-dimensional row vector x ∈  ,
(i ) (i ) (i ) (i ) (i )
x =  x1 x2 x3 x4  .
Each feature dimension is a 150-dimensional column vector x j ∈ 150×1 ,
for example:

 x j (1) 
 ( 2) 
 x 
xj =  j 
  
 x (150) 
 j .
Similarly, we store the target variables (here: class labels) as a
 y (1) 
 
150-dimensional column vector y =  …  ( y ∈ {Setosa, Versicolor, Virginica} ) .
 y (150) 
 

A roadmap for building machine learning

systems
In the previous sections, we discussed the basic concepts of machine learning and the
three different types of learning. In this section, we will discuss other important parts
of a machine learning system accompanying the learning algorithm. The diagram
below shows a typical workflow diagram for using machine learning in predictive
modeling, which we will discuss in the following subsections:

[ 12 ]
Exploring the Variety of Random
Documents with Different Content
taken nearly forty years to convince the world that tactics are as
necessary for the effective employment of machine guns as for
infantry. The general principles for their tactical employment have
already been dealt with in Chapter II., but they are so excellently
condensed in Sect. 187 of the German Regulations, that at the risk
of repetition they may be quoted here.
“Machine guns enable commanders to develop at fixed points the
maximum volume of infantry fire on the smallest possible front.
Machine guns can be employed over any country that is practical for
infantry, and when they are unlimbered they must be able to
surmount considerable obstacles. In action they offer no greater
target than riflemen fighting under like conditions, and they can, in
proportion to their fire value, support far greater losses than
infantry. They can utilise all cover that infantry are able to use.
Cover which is barely sufficient for a section of infantry (60 men) can
protect an entire machine-gun detachment (six guns).”
In order to see in more detail how machine guns should be
employed with infantry and the limitations of their tactics, it will be
necessary to follow the sequence of the infantry combat in attack
and defence, and to assume such situations as seem likely to arise.

ADVANCED GUARDS

The action of the infantry of an advanced guard will follow the

same lines whatever the dimensions of the force, and the battalion is
a convenient unit with which to deal. There are four battalions in a
brigade, and three brigades in a Division; consequently the
Divisional General will have six batteries of four guns each, or 24
machine guns, under his command. We will assume that the
battalion we are dealing with forms the advanced guard of its
brigade, and that the Brigadier has given it a battery of four machine
guns, two of which will probably belong to the battalion.
Field Service Regulations, 1909, page 79, says: “An advanced
guard is divided into a vanguard and a main guard.... The special
duty of the vanguard is reconnaissance. It will therefore generally be
composed of the advanced guard mounted troops, with or without a
body of infantry as a support. By day, when the country is open and
the advanced guard is strong in mounted troops, infantry will not as
a rule form part of the vanguard.... The main guard comprises the
troops of the advanced guard not allotted to the vanguard.”
We will first take the case of infantry with the vanguard as a
support to the mounted troops, and suppose that two companies are
allotted for this duty and have been given a section of machine
guns. The formation of the infantry of the vanguard will largely
depend on the nature of the country and the proximity of the enemy.
Where the country is open plain they will probably be extended in a
long line of skirmishers, in which case the machine guns should
march close in rear of the centre on the road. Should the country be
close or broken, the vanguard infantry may be confined to the road
—with small parties pushed out to the front and flanks—in which
case the machine guns should march in rear of the advanced party,
and not in rear of the vanguard. It must be remembered that
machine guns are able to come into action and open a heavy and
accurate fire in less time than it would take infantry of equal fire
power to deploy and to open fire from a position; it is therefore the
machine guns of the vanguard that should be the first to open fire
and give the infantry time to deploy and find fire positions. When
the enemy is encountered the object of the vanguard infantry must
be to support the cavalry as quickly as possible, and enable them to
mount and push forward round the flanks.
The section commander must go forward at once with scouts from
each gun, and quickly grasp the situation, consulting the officer
commanding the troops engaged as to the action to be taken; he
must then select the positions for his guns to come into action. The
scouts should take ranges and have everything ready for the guns to
open fire on arrival. The moment the guns open fire, the section
commander should send scouts off to find alternative positions to
the front and flank from which it may appear possible to enfilade the
enemy. Should the target be unsuitable, e.g. a line of well-concealed
skirmishers, etc., fire must be reserved, and the guns should take up
positions of observation while the section commander and scouts
seek for a position on the flanks from which to bring an effective
fire.
As soon as the vanguard infantry are deployed the section
commander must co-operate with them entirely, and must take his
instructions from the vanguard commander, who should leave him a
free hand, merely telling him his intentions and how the guns may
best assist him.
When the main guard comes up, in the case where the vanguard
is checked, the section commander should take the earliest
opportunity of placing himself under the battery commander, and at
this stage the guns should be used to seize and hold positions of
importance, especially artillery positions. If the advanced-guard
action discloses the fact that the main body of the enemy has been
encountered, and the G.O.C decides to engage, the advanced guard
will be required to seize all positions of tactical value and to hold off
the enemy until the main body has had time to arrive and deploy.
Their action will vary according to whether the G.O.C. decides to act
on the offensive or defensive. The latest German Regulations say:
“On becoming engaged, that side will have the advantage which
gains the start over its opponents in readiness for action, and thus
reaps the benefit of the initiative.” No arm is better able to assist in
this than the machine gun, and every available battery should be
brought up at this stage to take part in the advanced-guard
engagement.
The Germans say: “The advanced guard will fight on a wider front
than that ordinarily allotted to a force of its size, so as to seriously
engage the enemy. It will be disposed in groups more or less
detached, which will occupy those positions the possession of which
will most protect the deployment of the main body. It will be
assisted by detachments of machine guns, which will be placed for
preference on those positions which it is most necessary to prevent
the enemy from capturing.”[16]
 The commander of the machine guns of the advanced guard is
under the immediate orders of the O.C. advanced guard, and must
co-operate with him throughout the engagement. All officers with
machine guns must remember that isolated action is useless, and
that however effective their fire may be locally, unless it directly
furthers the object of the action, it is an unjustifiable waste of fire
power.
The seizure of ground likely to be of use to the main body—
particularly artillery positions—must be the first consideration of
machine guns with the advanced-guard infantry, once the infantry
are in action. That they can hold these unsupported is certain.
“There never has been and cannot be such a thing as a successful
attack on a line of machine guns in a favourable position in action,
until the machine guns have been well battered by artillery fire,”
says Lieutenant Parker in his book Tactical Organisation and Uses of
Machine Guns in the Field.
In the cases where the infantry of the advanced guard forms the
main guard, the battery of machine guns should march with the
foremost body of troops and should never march in rear, where they
are useless, and where, if on a road, they can only be moved up
with difficulty. The battery commander should remain with the O.C.
advanced guard until he is fully acquainted with the situation and his
intentions. Should the situation be such as to enable the advanced
guard to brush aside the opposition, the machine guns must be used
vigorously to support the infantry attack, and should be pushed
forward to seize positions before the infantry deploy, and cover their
deployment by fire. In the preliminary stages these positions will
probably be on the immediate front, and the four guns should be
able to occupy a front roughly coinciding with the first deployment of
the infantry. As soon as the infantry come into action the machine
guns may be withdrawn and used on the flanks to bring a cross fire
to bear, and to fire on any closed bodies, led horses, etc., that may
offer a target. At this stage the guns should work in pairs in mutual
support, and may be used on one or both flanks as the situation
requires.
 Machine guns should rarely be used in the firing line after the
infantry deploy, as they offer a rather marked target and can be far
more profitably employed in pressing the attack on the flanks.
There are several instances of the successful use of machine guns
with the advanced guard in South Africa in 1900 to support the
mounted vanguard, and on many occasions they were able to brush
the enemy aside without deploying the main guard or delaying the
march of the column.

THE ATTACK

Under cover of the advanced-guard action and the subsequent

artillery duel the infantry will deploy for the attack. The first advance
of the infantry will probably be covered and supported by the
artillery, and they will generally be able to push forward to within
long range of the enemy’s infantry, without the necessity of covering
their advance by rifle fire. From here, however, the firing line will
begin to suffer from rifle fire, and it will be necessary to reply to this
fire in order to cover the further advance.
“To enable the attack to be prosecuted it is clear that from this
moment the hostile fire must be met by fire of greater power, its
intensity and duration being dependent on the effect produced.
Moreover, from this point forward movement will as a rule be
contingent on the mutual fire support afforded to one another by
neighbouring bodies, and on the effects produced by the covering
fire brought to bear on the enemy’s line wherever the ground admits
of such action, by supports, reserves, and the troops specially told
off for this purpose. Covering fire should be delivered so as to strike
the enemy, disturb his aim, and compel him to seek shelter when
the assailant is exposed or in difficulties. In such circumstances
intensity of fire is of the first importance; but fire should be reserved
when the assailants’ progress is satisfactory, or when he is under
cover.”[17]
 No one who has controlled the covering fire of infantry, even on
manœuvres, will dispute the immense difficulty of timing the
opening and ceasing of such a fire, with even a company, to coincide
with the advance of the troops covered; while the necessity for
observation of fire and its concentration on certain parts of the
enemy’s position make it still more difficult effectively to cover the
advance of troops with rifle fire at long range.
That machine guns will be able to perform this duty far more
effectively and easily than infantry is manifest from the nature of the
fire required, the features being intensity, control, and
concentration, all of which are characteristic of machine guns. When
required to cover the advance of infantry, machine guns should be
used in batteries and must be so placed that they are able to see the
infantry they are covering, and to sweep those portions of the
enemy’s position from which fire can be brought to bear on the
advancing troops. If the enemy’s artillery is not silenced or at least
dominated by that of the attackers, it will be necessary to conceal
the machine guns from view and provide cover from fire. That
machine guns well concealed and provided with cover (pits) are able
to remain in action under artillery fire for considerable periods, is
proved by instances in the Russo-Japanese War (see Chapter III. p.
28). Major Kuhn, who was with the Japanese Army in Manchuria, in
his report states: “Importance is laid on concealing the guns, and it
was claimed that none had been knocked out by the Russian
artillery.” Covering fire from the flanks is more likely to be effective
than frontal fire, because it will enfilade entrenchments, reach men
behind natural cover and have a more disconcerting morale effect
than frontal fire. It will not always be possible to find suitable fire
positions on the flanks for machine guns at this stage of the attack,
and it will then be necessary to place them in rear of the attacking
infantry and fire over their heads. This can be done with perfect
safety—indeed, there is far less danger of accidentally hitting the
attacking infantry with machine guns than with artillery fire. Direct
fire should only be used when it is impossible to use indirect fire on
account of the nature of the ground in the vicinity.
 Indirect fire from the reverse slope of a hill has the advantage of
concealing the guns and rendering them immune from artillery fire.
The position must be carefully selected with due regard to the slope
in its relation to fire from the enemy as laid down in chap. ii. p. 67,
of Training Manuals Appendix, 1905. The best position is just below
the crest of a steep slope, as observation of fire is rendered easier
and the effect of shrapnel minimised. The safety of the infantry in
front will depend entirely on the range, height of the guns above the
infantry, and the position of the target (enemy). On level ground the
range must not be less than 800 yards,[18] and the infantry will be
perfectly safe at all distances between 200 yards from the guns and
200 yards from the enemy. (See table, Appendix A.) The methods of
employing indirect fire have been given at the end of Chapter II.,
and its success will depend largely upon the careful observation of
fire by the section commanders, who must concentrate their
attention on this alone—the battery commander giving the signal for
opening and ceasing fire, and the target or direction of fire for each
section. Machine guns used as covering fire have ample time for
preparation; and as ranges can be accurately taken, and great
deliberation used in selecting positions, observing fire, and
correcting errors, the maximum of accuracy should be attained. If
the enemy is entrenched against frontal fire, good results may be
expected from accurate distant fire, owing to the angle of descent of
the bullet causing the low parapet or shallow trench to be no
protection.
They were successfully used by the Japanese as covering fire for
infantry in the attack, as the following examples will show: “At
Mukden on March 1st all the machine guns of a whole Japanese
division (12 to 18 guns) were brought into action upon a Russian
point d’appui. The Russian fire was silenced, but burst out again
whenever the machine-gun fire slackened. The Japanese infantry
used these pauses in the enemy’s fire to press forward to close
range under cover of their machine-gun fire.”[19] On March 2nd the
three machine guns of the 10th Japanese Infantry Regiment acted in
the same way against a Russian fieldwork. This method of
employing machine guns requires the closest co-operation with the
infantry from the commencement of the attack.
Again, during the Japanese attack on Namako Yama the infantry
were greatly assisted by covering fire from their machine guns
directed on the Russian trenches. These guns were used from
behind screens, and their success was largely due to their being well
concealed.
It will rarely be advisable for machine guns to follow infantry into
the firing line, where they present a conspicuous target which
attracts fire and renders their withdrawal difficult.
We made this mistake in the South African War more than once.
At Rietfontein the machine-gun detachment of the Gloucester
Regiment, which had followed the battalion into the firing line, was
almost annihilated. At Modder River the Scots Guards Maxim gun
accompanied the firing line, and the detachment was annihilated by
pom-pom fire, and the gun was left on the field alone all day.[20] In
the attack on Cronje’s laager at Paardeberg, machine guns were
used in the firing line on the left bank of the river, and when the
attack failed the machine guns, having suffered severe losses, could
not be withdrawn and had to be abandoned till nightfall.
When the covering fire of machine guns is no longer considered
necessary, they should be withdrawn and concentrated in batteries
in rear of the reserve or in such other convenient position as the
G.O.C. may direct. They should take this opportunity of refilling
belts, replenishing ammunition, water, etc., and if the guns have
fired many thousand rounds, of exchanging barrels.[21] They are
now at the immediate disposal of the G.O.C. and will be used by him
as a mobile reserve. Circumstances vary so in war that it is
impossible to particularize in their use at this stage, but their great
mobility will render them extremely valuable in the following cases:
1. To assist a turning movement.
2. To reinforce a distant flank.
 3. To repel a counter-attack.
4. To hold a captured position.
When the infantry reach close range, the point for the assault will
have been selected and the reserves massed behind this point.
When the reserves have been thrown in, and the fire fight has
reached its height, the moment will have arrived when one side or
the other will obtain superiority of fire. “The climax of the infantry
attack is the assault which is made possible by superiority of
fire.”[22] To assist in the attainment of this superiority of fire is the
true rôle of machine guns with infantry, and, as their beaten zone at
short ranges is so small, they can safely fire over the heads of prone
infantry, even when the latter are within a hundred yards of the
target. The guns should now be pushed up to the closest ranges,
and fire of the most rapid description concentrated on the point of
assault. When it is remembered that the machine guns of a single
Infantry Division, as at present organised, can bring a fire of more
than ten thousand shots a minute from close range where its
accuracy is assured, at this crisis, there will be no need to say any
more on the tactical importance of the probable result.
In bringing machine guns into action for this purpose, they will
generally be used in batteries, but may come into action in sections
or even single guns once the objective has been pointed out. The
object being to obtain superiority of fire regardless of cost, rapidity
and concentration of fire must be the principal points. The guns
must therefore press forward to the closest range, and where it is
possible to bring a cross or enfilading fire to bear, the opportunity
must not be neglected. It may be necessary to fire over the heads of
the infantry or even to push guns into the firing line itself, though
this is seldom desirable if good positions can be found in rear. When
any portion of the line advances to the assault, fire must be
concentrated over their heads on the position assaulted; and when
they are so close to the position as to render such fire impossible,
fire must not cease, but be directed over the position so as to strike
the enemy as they retreat. Fire from machine guns may be kept up
from 500 yards’ range over the heads of prone infantry within 100
yards of the target with perfect safety, and from 800 yards over
infantry advancing to within the same distance. General Nogi,
speaking of the use of machine guns in the attack, says: “Our troops
trained machine guns on the most advanced lines of infantry to
overwhelm with fire the points at which resistance was greatest.
Sacks of earth were used to mask them. They have often enabled
the infantry to advance with success.” M. Ullrich, war correspondent
of the Gazette de Cologne, was present in many engagements in
which machine guns were pushed up to support the firing line in the
final stage. He says: “In the offensive the Japanese frequently made
successful use of machine guns. When the infantry were carrying out
a decisive attack, they were supported by their machine guns, which
concentrated their fire on points arranged beforehand.... When
machine guns have been skilfully employed, their action has been
infinitely more effective than that of field artillery, more especially
when they fire at infantry ranges.” The following is an example of
their actual use in this way. “The Japanese,” says a Russian eye-
witness of the fighting round Mukden, “brought up during the night
dozens of machine guns with hundreds of thousands of cartridges to
their front line of skirmishers, from 400 to 500 yards from our
positions, and entrenched them there. When the assault
commenced, at dawn, the machine guns opened fire with fatal
accuracy on the parapets of our trenches and on our reserves,
preventing them from coming up. We could do nothing with the
enemy, because when the machine guns showed the least
vulnerability, they were at once protected by shields of bullet-proof
steel.”[23]
The machine guns of the enemy will all be in action during the
final stages, but they should not be engaged by the machine guns of
the attack except when exposed. They are very difficult to put out of
action by small-arm fire at any time, and when employed by the
defence are sure to be well concealed and provided with cover. It is
the duty of the artillery to silence machine guns, and this was so far
recognised in the war between Russia and Japan that the latter
brought up mountain guns on more than one occasion to silence the
Russian machine guns.
At Kinsan on June 26th, 1905, when the Japanese were attacking
the position, the Russians brought up two machine guns against the
43rd Regiment and a mountain battery at 3 p.m. The battery at once
silenced the machine guns, and by 5.30 p.m. the hill was in the
hands of the Japanese.[24] Captain Niessel, of the Russian Army,
gives another instance. “It was on August 31st at 7 p.m. that the
Japanese decided to drive the machine guns out of the village of
Goutsiati by artillery fire. Knowing that their infantry had been
checked, they posted at the village of Datchaotsiati a battery which
showered on us shrapnel and high explosive shell. Although the men
had taken cover behind a wall of earth, we had many wounded and
could not breathe freely until nightfall, when the enemy ceased their
fire, to which we could not reply on account of the distance. At 9
p.m. I was ordered to evacuate the position.”[25]
The moment the infantry assault is successful, the machine guns
must be pushed forward into the captured position to secure the
ground gained and to repel counter-attacks. As the confusion at this
moment will be considerable, it will seldom be possible for the
machine-gun commander to receive orders from the G.O.C., but this
must not prevent him from immediately directing guns to occupy
and quickly entrench themselves in positions which it may appear
vital to secure, and to order the remainder of the guns not so
required to follow and fire on the enemy wherever he may show
signs of rallying. In order that these two distinct duties may be
carried out instantly and without confusion, it will be advisable to tell
off beforehand those batteries that are to pursue the enemy.
Theoretically every available gun should be launched in the
pursuit, but the recent war between Russia and Japan has shown
how often a position that has been stormed and captured has been
retaken by a rapid counter-attack before the attackers have had time
to rally and prepare to hold what they have taken. Machine guns can
render a position safe from counter-attack so quickly and effectually
that this duty should be the first consideration of a machine-gun
commander with infantry in the assault.
Fresh ammunition must be brought up, and every endeavour
made to conceal the guns and construct good cover from any
material found in the position.
“During the battle of Mukden, on the evening of March 1st, a
position at Sha-shan was captured by the Japanese. The Russians
had taken up another position some 500 yards in rear, from which
they opened an effective fire upon the Japanese infantry in the
captured position. Another Russian force about 1,000 yards further
to the west threatened their flank. The Japanese, however,
succeeded in bringing up the machine guns which had taken part in
the attack, and brought them into action under cover of a number of
sandbags abandoned by the Russians. The effect was decisive: all
counter-attacks failed before the annihilating fire of the machine
guns.”[26]

IN THE DEFENCE

When infantry are acting on the defensive, it is advisable that the

majority of the machine guns should be held in reserve, and used to
check turning movements, to reinforce distant portions of the line of
defence, and to deliver counter-attacks, but their principal rôle will
be the repelling of the assault. Guns allotted for this purpose must
be placed in very carefully selected positions in the line of defence,
from which they can command narrow approaches, and sweep
ground in front of trenches or important works by cross fire,
especially any dead ground likely to afford temporary cover to the
attackers. Concealment is of the first importance, and cover from fire
absolutely necessary, but the combination of the two requires great
skill both in choosing positions and in constructing an inconspicuous
shelter. In certain positions where the country is flat or undulating
and has no abrupt features, pits will be found the best form of
cover; but where the ground presents marked features,
emplacements providing overhead cover will not only conceal the
guns, but render them immune from shrapnel. The Japanese used
these emplacements with great success, the guns often remaining
concealed even when the enemy were within 300 yards.
The positions for machine guns should not be in one and the same
line, and they should be placed at wide intervals. At least two
positions should be provided for each gun, and a second series of
positions in rear, commanding those in front, should also be
provided. The machine guns should be allotted to their positions by
sections, each gun being so placed that its fire crosses that of the
other when possible. Section commanders should be responsible for
having all ranges from their respective positions taken and written
up in each emplacement. They must arrange with the O.C. of that
portion of the position to which they are allotted the time and signal
(if any) for opening fire; the exact moment should be left to the
discretion of the section commander as a rule, the O.C. merely
indicating the stage of the attack up to which fire is to be reserved,
and great care should be taken not to open fire before this stage has
been reached. Every endeavour should be made to ensure co-
operation between the machine-gun units in reserving fire till the
shortest range has been reached by the attack and a really good
target presented. Nothing can justify a machine gun opening fire
from a concealed position until its fire will give the best results that
can be expected from that place, and to ensure this it is necessary
that the target is big and vulnerable, the range close, and the fire
unexpected by the enemy. Captain von Beckmann, speaking of the
Russo-Japanese War, says: “Premature fire upon unsuitable targets
at long range is to be deprecated. The greater the surprise, and the
shorter the time within which heavy loss is inflicted, the greater the
morale effect produced. At the battle of Hai-kou-tai on January 27th,
1905, a Japanese company attacked Sha-shan. Four Russian
machine guns opened fire at about 1,100 yards upon the extended
firing line, without causing any serious loss or affecting its advance.
On the other hand, the Japanese on March 1st had approached to
within 200 or 300 yards of the Russian position at Wang-chia-wo-
pang and were beginning the final assault. Two Russian machine
guns suddenly came into action, and the Japanese assault was
repulsed with heavy loss owing to the annihilating fire.”
These two instances well illustrate the right and wrong way of
using machine guns in the defence. Where a battery is allotted to
one section of the defence the guns should be distributed two or
four along the front and one or more on each flank. Single guns may
be used in the defence where a section cannot be spared, as the
assistance of a supporting gun is not essential to its safety when
behind entrenchments. In the first instance guns may be kept in rear
of the position until the direction of the attack has been ascertained,
provided they are able to occupy their positions unseen.
Machine guns at salients and on the flanks are of great value in
defending a position.
If sufficient machine guns are available, one or two batteries
should be reserved for the counter-attack. They should in this
instance be pushed right into the firing line when necessary and
used with the greatest boldness. It will be their duty to cover the
retirement of the Infantry in the event of a repulse, and for this
purpose positions in rear should have been selected beforehand.
Although the principle of reserving machine guns in the defence
must be strictly adhered to, there may be occasions when the use of
a battery or two well in front of the defensive position will compel
the enemy to deploy prematurely, cause him loss, and delay his
advance. Machine guns so used may also find opportunities for
surprising artillery while unlimbering, and thus render great service
to the defence. They should be withdrawn when their mission is
accomplished, and be reserved for further use as already indicated.
The following are examples from the Russo-Japanese War of the use
of machine guns in the defence of a position.
General Nogi has written: “Our most formidable foes were wire
entanglements placed 100 yards in front of the Russian trenches,
well lit up by search-lights and covered by the murderous fire of
machine guns. The defenders made use of them to flank dead
ground in their front and also had them at other points, kept
carefully in reserve and under good cover, to make use of continuous
fire against the attackers at the moment of the assault.”
“On January 28th, 1905, near Lin-chin-pu at about 7 p.m. the
Japanese attacked the forts of Vosnesenski and the trenches near
by, in which were posted two machine guns. These latter opened fire
at 200 or 300 yards on a Japanese company in line. In one or two
minutes they fired about 1,000 rounds, and the Japanese company
was annihilated.”[27]
“At Mukden on March 1st, the left of a Japanese Division being
within 300 yards of the enemy’s position and about to assault, the
Russians suddenly opened a very heavy machine-gun fire from
cleverly concealed positions, causing such loss that the Japanese
attack was temporarily suspended. On August 20th, 1904, the
Japanese captured a lunette near the village of Shin-shi after severe
fighting. The Russians made a counter-attack with three machine
guns, and drove the Japanese out again with a loss of over 300. The
three machine guns retired from the lunette before the attackers got
home, and, taking up a position behind the open gorge of the work,
showered such a hail of bullets on the victorious Japanese that they
were compelled to retire.”[28]
“On February 27th, 1905, the Russians tried to surprise the
railway bridge over the Sha-ho on a very clear night, when one could
see as far as 500 yards. Four Japanese machine guns opened fire on
a company of the 10th Light Infantry, which was almost wiped
out.”[29]
At the battle of Hei-kow-tai, the Russians at Shen-tan-pu made no
less than five determined attacks against the Japanese
entrenchments, in which was a machine gun, but were repulsed
each time. The machine gun is said to have done great execution,
and one thousand dead Russians were reported to have been found
before it.[30]
 The 8th Division of the Japanese Army are reported to have made
several fine attacks upon Hei-kow-tai, but were each time repulsed
mainly by the fire of the Russian machine guns.[31] The same report
says, “Throughout the campaign in Manchuria the Japanese have
suffered severely in attacking those points of the Russian front which
have been armed with machine guns,” and proceeds to quote an
episode in the operations of the 5th Division at the battle of Mukden,
to show to what lengths the Japanese went in order to silence these
weapons. It appears that the Japanese were so galled by the fire of
four machine guns that the attack was materially affected. They
decided to bring up two mountain guns to within 500 yards of these
machine guns, to try to destroy them. The mountain guns were
brought up behind a wall and fired through two holes bored for the
muzzles. Two machine guns were at once destroyed, but so well had
the others been concealed that they were able to be withdrawn in
safety.
Sufficient has been quoted to show that both the Japanese and
Russians made the greatest use of machine guns in the defence, and
that when employed on sound tactical principles they not only
afforded material assistance, but were often the predominant factor.
On the other hand, when these principles were neglected or ignored,
the machine guns merely wasted ammunition and were impotent to
affect the situation. The lesson to be learned is this: that machine
guns are only useful when their tactical handling is thoroughly
understood, and then their effect is more decisive than that of any
other arm.
 CHAPTER VI
EMPLOYMENT IN THE FIELD WITH INFANTRY (contd.)

THE RETREAT

The reply of the British drummer-boy when asked by the great

Napoleon to beat the “Retreat,” is historical, and, following the
traditions of the British infantry, the word will not be found in the
index of Infantry Training. Nevertheless, although British infantry
never retreat, they may be called upon to perform that most difficult
operation of war—“a retirement in the face of the enemy.”
Field Service Regulations, Part I., 1909, speaking of the retreat
says: “The cavalry and other mounted troops, aided by a strong
force of artillery, will meanwhile check the enemy’s advance, and the
remainder of the force, with the exception, if possible, of a rear
guard in support of the mounted troops, will move as rapidly as
possible to the shelter of the rallying position and there reorganise.
Steps should be taken immediately to secure any bridges, defiles, or
other vital points on the line of retreat at which the enemy’s
mounted troops might intercept the force. Commanders of retreating
troops must recognise that their greatest danger will arise from
attacks in flank delivered by the enemy’s mounted troops and horse
artillery; if possible, therefore, they should take precautions that all
ground commanding their line of retreat is occupied by flank
guards.”
There are three distinct ways of employing machine guns with
infantry to assist in covering the retreat, as will be seen from the
portions of the above quoted in italics, viz. (1) With the rear guard.
(2) To secure vital points on the line of retreat. (3) To occupy
commanding positions on the flanks. The use of machine guns with
the rear guard will be dealt with separately, so we will see how they
should be employed to fulfil the two latter duties.
Let us suppose that the moment has arrived when the commander
of the force decides to break off the engagement and to effect a
retreat. We have already seen how the cavalry and their machine
guns will co-operate, and we may assume that their intervention will
enable the infantry to break off the fight under cover of the fire of
the rear guard and their machine guns. In anticipation of this, the
commander of the machine-gun batteries—other than those allotted
to the rear guard—will assemble them in a central position and
receive from the commander of the force his plan of action and
instructions for the co-operation of the machine guns. These
instructions should definitely state the positions to be occupied on
the line of retreat, and for how long such positions are to be held
(e.g. until the infantry are past such-and-such a place, or to the last
possible moment, etc.); but they should leave all details as to unit of
guns, fire positions, opening fire, etc., to the machine-gun
commander, who should be given a free hand in carrying out his
instructions. The machine-gun commander will then issue his orders
to battery commanders, allotting the battery for the flank and the
battery which is to occupy positions on the line of retreat. In like
manner the battery commander will tell off sections to occupy
certain positions, if these have already been selected, and if not he
will direct his battery to a position of readiness and will gallop off
with his section commanders to select the positions for each section
and explain the course of action to be taken. The position of
readiness must be in the vicinity of the position to be occupied, and
section commanders must take steps to ensure that their sections
keep in signalling connection with them and can be brought up by
signal to the position selected.
The choice of the position will depend upon the nature of the
country and the features favourable to delaying the enemy, but it
may be laid down that the machine guns should not be placed in
one line, but should be broken up as much as possible, so that no
two guns are at the same range. In the same way, guns placed in
well-concealed positions on the lower slopes of hills should have the
other guns of their sections higher up, where they can command a
wide extent of country and at the same time cover the retirement of
the guns below. All guns must be so placed that they can be retired
under cover of the ground immediately they move. The guns on the
higher ground will open fire at long range, and the opportunity may
be afforded at this stage for using “deliberate” fire at the most rapid
rate possible, in imitation of rifle fire, for the double purpose of
concealing the guns and deceiving the enemy by causing them to
believe that the position is held by infantry. The guns at the lower
position should watch their opportunity for opening fire on a large
and vulnerable target; but should none be presented, it is possible
by reserving fire to catch the enemy in close formation should they
press on when the guns have ceased fire, thinking the position has
been vacated.
It may be possible to surprise the enemy from close range at
bridge-heads or on roads through defiles leading to the line of
retreat, if guns are carefully concealed in positions commanding
such approaches. Indeed, the opportunities for ambushing pursuing
cavalry or infantry will be frequent during the early stages of a
retreat, and no chance must be lost in making the enemy pay dearly
for his temerity during the pursuit. One successful ambush by
machine guns is worth more than the most stubborn stand, for it
imposes caution as nothing else can; and caution in the pursuit
means delay, and delay spells failure. When machine guns get a
really good target at close range, the result is little short of
annihilation, and the morale effect of such a heavy loss, totally
unexpected and inflicted in a few moments, takes the life out of the
pursuit more completely than even a long stand.
The machine guns allotted to the flanks will find positions in the
way already described. These positions must be chosen along the
line of retreat so as to hold all ground from which fire can be
brought to bear on the retreating columns, and from which the
enemy can be successfully repulsed when he attempts to break
through. If there has been time to form infantry flank guards, the
machine guns will assist in strengthening their resistance, and in
covering their withdrawal. When acting alone they should open fire
on suitable targets at the longest ranges, but when supporting
infantry they may be able to reserve fire and surprise the enemy
from close range. Guns working together on a flank must keep in
touch with each other by signal, and guns should work in sections in
mutual support. A battery acting as a flank guard to a force
retreating must be prepared to occupy a front of some two or three
miles, and sections will often be in positions a mile apart in open
ground. Directly the guns have taken up positions the battery
commander should go off with scouts from each gun and select
fresh positions along the line of retreat. He will give the signal for
retirement, and if in action at the time one gun of each section
should move off first under cover of the fire of the others; or where
the situation permits, a flank section may be ordered to move first.
It must sometimes occur that the cavalry on the flanks will be
compelled to retire by a superior force, and when this is the case
machine guns in position may be able to inflict considerable loss,
should the pursuing cavalry follow them too closely while retiring.
As soon as it is known that the main body has reached the rallying
position and is ready to hold it, the machine guns can be withdrawn.
This will not usually be possible in daylight, and as a rule it will take
place after dark. The dispositions of the machine guns in the rallying
positions will be the same as in the case with infantry in the
defence.

REAR GUARDS

The prime necessity for the machine guns with a rear guard of a
retreating force is mobility, and without this they will be of little
service and become a hindrance rather than a help.
“A rear guard carries out its mission best by compelling the
enemy’s troops to halt and deploy for attack as frequently and at as
great a distance as possible. It can usually effect this by taking up a
succession of defensive positions which the enemy must attack in
turn. When the enemy’s dispositions are nearly complete, the rear
guard moves off by successive retirements, each party as it falls
back covering the retirement of the next by its fire. This action is
repeated on the next favourable ground.... A rear guard may also
effectually check an enemy by attacking his advanced troops as they
emerge from a defile or difficult country.”[32]
Machine guns, by reason of their concentrated fire and shallow
beaten zone, are peculiarly suited for compelling troops to deploy at
long range. They are also suited for quickly taking up a defensive
position, and, when mobile, can instantly cease fire and move off in
a few moments with considerable rapidity.
Their power for suddenly developing an intense fire will enable
them quickly to overwhelm an enemy caught “emerging from a
defile or difficult country.” It would appear therefore that the
machine guns are far more fitted than infantry to perform the duties
of a rear guard as quoted above.
“The first consideration in selecting a position for the artillery is
that it shall be able to open fire on an enemy at long range, and
thus compel his infantry to assume an extended formation at the
greatest possible distance. The second is that it should be possible
to withdraw without difficulty.”[33]
When the infantry arrive within effective rifle range of the artillery,
the latter will be compelled to retire, especially if the former are
supported by cavalry. This will also be the moment when the infantry
will require to retire to a fresh position in rear. Machine guns that
have been well concealed, or which have taken up fresh positions in
good cover in rear of the infantry line, should be able to unmask and
by their fire hold off the enemy until both guns and infantry have
effected their retreat. The facility with which they can traverse and
sweep widely extended lines, and instantly concentrate fire on any
portion of it, make them of great service in covering a retirement at
effective range, as they can render the hottest fire ineffective and
aimed fire impossible for a short time.
Machine guns with a rear guard must be prepared to work from
place to place with great rapidity, and for this purpose they must
arrange to bring the gun horses close up to their fire position.
Alternative positions are essential, but must be reached under cover
and unobserved by the enemy. Sections will act independently, using
their guns in mutual support; but the retirement of one section may
often afford another the opportunity, from a well-chosen position on
a flank, to surprise the enemy from close range as they press
forward to occupy the abandoned position.
A ruse that may sometimes be successful where guns are really
well concealed is to bring up the gun horses and retire at a gallop by
a pre-arranged signal all along the position, but leaving one gun of
each section still in position but flat on the ground, where it will be
completely concealed even from glasses. This is almost certain to
draw a hot pursuit, especially if the artillery and infantry have retired
previously, and if fire is reserved till close range a severe check may
be administered to the enemy. “Skilfully laid ambushes will cause the
enemy to move with caution in pursuit.”[34]
Machine guns should never retire for a few hundred yards, except
where absolutely necessary for covering the retirement of those in
front. Once in position, they should only be moved to alternative
positions when discovered, and these will usually be more or less on
the same general alignment. When they retire they should move to
such a distance in rear as to give them time to select, or if necessary
improvise, good cover in the new positions and to replenish
ammunition, etc. “The positions should be sufficiently far apart to
induce the enemy, after seizing one, to re-form column of route
before advancing against the next.”[35]
General Alderson, speaking at the Aldershot Military Society in
1904 and referring to the South African War, said: “I had two
Maxims with tripod mountings on pack-saddles, which belonged to
the 1st Battalion Mounted Infantry. These guns had well-trained
mounted detachments, and a pushing officer, with a good eye for
ground, in command. They were most useful, and more than once
saved the flank of their unit from being turned by galloping up and
coming into action on the flank of the out-flanking Boers.... I am of
opinion that if the most is to be got out of the guns, the
detachments, even with infantry, should be mounted.... With the
detachment mounted ... there is no question about the guns not
keeping up, and they can be sent quickly to any desired position.
They can hold on to any such position as long as required to cover
the advance or retirement of their infantry, and then easily catch
them up or get into another position. In fact, if the detachment is
mounted the value of the guns is more than doubled.”
Machine guns with the rear guard will certainly be exposed to
artillery fire, and they will seldom have the time or opportunity for
making sufficient cover to protect themselves. It will be necessary
therefore either to withdraw the gun on coming under artillery fire or
to find shelter from it. If good natural cover has been obtained, it
will only be necessary to lower the gun flat on the ground behind it,
and for the detachment to lie flat close against it, when they will be
practically safe. The artillery will cease fire as soon as they find there
is nothing to fire at, but the detachment should not move for some
time after this, as the guns, having got the range, will be able to
inflict considerable damage if the detachment exposes itself.
It is always advisable to keep one or two sections in reserve with
a rear guard to be used to protect the flanks or any point where the
pursuit threatens to break through in the manner mentioned by
General Alderson. These sections should remain with the rear-guard
commander and be under his immediate command; they should be
in readiness for instant action.

OUTPOSTS

The duties of the outposts are:

(1) To provide protection against surprise.
 (2) In case of attack, to gain time for the commander of the force
to put his plan of action into execution.[36]
The first duty of outposts, which is reconnaissance, cannot be
assisted by machine guns; but their second duty, which is resistance,
may be materially strengthened and aided by them, as we have
already seen in dealing with the defence. Little can be said about
tactics, and it will be necessary to use the guns in positions on the
line of resistance where their fire power is most likely to be of value.
In order clearly to understand where and how machine guns should
be used with outposts, it will be necessary first to examine the
composition and distribution of outposts of a force of all arms. Field
Service Regulations, Part I., 1909, says: “When there is any
possibility of a force coming in conflict with an enemy, its
commander, when halting for the night, should first decide on his
dispositions in case of attack, and then arrange the quartering of his
command and the position of the outposts accordingly.... Command,
co-operation, and inter-communication will be facilitated by placing
the advanced troops along well-defined natural features, such as
ridges, streams, the outer edges of woods, etc., or in the vicinity of
roads, but this must not be allowed to outweigh the necessity for
making the best tactical dispositions possible. In enclosed country,
and at night, the movements of troops are generally confined to the
roads and tracks, which should be carefully watched. If the outpost
position is extensive, it may be divided into sections, each section
being numbered from the right. The extent of a section depends
upon the amount of ground which can be supervised conveniently by
one commander. The extent of frontage to be allotted to each
company will depend on the defensive capabilities of the outpost
position, and, where they exist, on the number of approaches to be
guarded. The outpost position will invariably be strengthened and
communications improved where necessary. Piquets and supports
will do this without waiting for definite orders.... The outposts of a
force of all arms consist of outpost mounted troops, outpost
companies, and, where necessary, the reserve. Machine guns will
generally be included and sometimes artillery.... When stationary,
the duty of local protection will fall almost entirely on the infantry,
most of the outposts mounted troops being withdrawn, their place in
this case being with the reserve, if there be one. In certain cases,
however, standing mounted patrols may be left out at night with
advantage....
“Machine guns with outposts may be employed to sweep
approaches, and to cover ground which an enemy in advancing may
be compelled to pass or occupy.”[37]
Machine guns must be used on the principles given above, which
naturally fall under two distinct headings, viz. use by day and use by
night. Machine guns with outposts by day will not occupy their fire
positions unless an attack is imminent, but should be posted with
the reserves in their own section of the defence. This must not
prevent emplacements or pits for the guns being made, ranges
carefully taken and written up, and everything being ready in case of
action. These positions and the subsequent action will correspond so
closely with those already suggested for infantry in the defence in
the last chapter, that they need not be repeated. By night, however,
the case is very different, and it is necessary to select very carefully
the position of each gun, so that it commands a road, a bridge, or
other ground which the enemy in advancing may be compelled to
pass or occupy. These places must be selected in order of
importance, and an endeavour made to leave no approach by which
a body of troops might move uncovered by fire. The machine guns
must occupy their positions before dark and be carefully laid to
sweep the area of ground necessary for protecting the position, the
angle of elevation of the gun being taken by clinometer afterwards
in order that it may be relaid if necessary in the dark. Great care
must be taken to conceal these guns from observation when
bringing them into position or withdrawing them, and it is worth
while going to some considerable trouble to do this. One method of
many is suggested to show what is meant. Field guns might be
placed in the positions by day and withdrawn before night, the
machine guns being brought up on the limbers and dropped in the
position behind a screen of brushwood, and withdrawn again in the
morning when the field gun is brought up.
The detachment of each machine gun must find two double
sentries and their reliefs, who will remain at their posts with the gun,
which should have the belt ready in the feed block, but not actually
loaded. These men may either take it in turn to stand to the gun or
both be directed to watch. When there are no infantry sentries on
piquet duty in front of the gun, it will be necessary to detail two who
will be posted in the usual way some little distance in front of the
gun to prevent it being rushed in the dark. The method of preparing
the sights and laying the gun for night work is given in Chapter VII.
The value of machine guns with the outposts at night is shown by
an incident at the Battle of Mukden, March 6th, 1905. Two Russian
battalions made a night attack against the hill north of Tung-chia-
wen, which was occupied by the 2nd Japanese Regiment. There was
no moon and the night was very dark. Two Japanese machine guns
did great execution at ranges between 50 and 100 yards, and the
Russians were repulsed with a loss of 450 men, the Japanese
casualties being only 48.
 CHAPTER VII
EMPLOYMENT IN FORTRESS WARFARE

The siege of Port Arthur by the Japanese and its defence by the
Russians have thrown an entirely new light on Fortress Warfare.
Although the main principles of the attack and defence remain
unchanged and the primary armament of both is still the heavy
artillery, the improvement of small-arms, their great range, and rapid
fire have materially altered the nature of the fighting in the later
stages and during the assault. The result of this has been to prolong
the defence after the fortress has been dismantled by the
bombardment, and to render the storming of a single breach an
operation no longer possible in war. The machine gun is largely
responsible for this; and when high-explosive shells have destroyed
the fortifications and disabled the artillery, the stormers will be met
by rifle fire and that of machine guns concealed among the ruins of
the works. The intensity and accuracy of this fire will be such as to
result frequently in the complete repulse of the assault, and even
when the glacis of a work has been occupied it may be several
weeks before the capture of the work itself is effected.
Machine guns are particularly suited to the defence of fortresses
during this period, and Sir G. S. Clarke, in his book Fortification,
says: “The fire of the Maxim gun, delivering about 700 bullets a
minute,[38] can be directed by one man, who need not show more
than his head (easily shielded) above the parapet, the feed being
tended by another man completely under cover. In the special
qualities of the machine gun there is a distinct advantage to the
defence, arising from the fact that an intense fire over a particular
area can be suddenly developed by a few men occupying a small
space. This, in the case of night attacks especially, is a valuable
quality. At Port Arthur the Russians in some cases employed machine
guns with good effect, concealing them so that their fire came as a
surprise to the assaulting parties. Their portability renders them well
suited for the defence of positions, and they will doubtless form an
important element in the armament of fortresses.”
Speaking of the difficulty of “storming” the modern fortress, the
same author says: “The vive force school proposed therefore—on
paper—to shell them heavily and then storm, trusting to incomplete
organisation and general unpreparedness. There is little or nothing
in military history to bear out the views of this school, and modern
experience is entirely against them. Only one such attempt was
made in 1870-1, against the indifferent provisional works of Belfort,
garrisoned mainly by Gardes Mobiles—and this failed completely.
The tremendous assaults on the defences of Port Arthur may have
been partially inspired by the German teaching; but the results were
discouraging, although the devoted and sustained gallantry of the
Japanese could not be surpassed and probably would not be
equalled by any European army.”
The machine guns of a fortress should be divided into two classes
—the stationary and the mobile guns. To the former will be allotted
the defence of distinct portions of the permanent works, and they
will be provided with cone and parapet mountings,[39] the former
being fixed and the latter capable of being moved within the work to
which it is allotted.
The mobile guns should be mounted on a light tripod and carried
in a low-wheeled handcart, or they may be mounted on a very light
two-wheeled carriage capable of being drawn by one man and
having wheels of small track, which can move over the narrowest
roads in the fortress. These mobile guns should not be allotted to
any particular work, but to the garrison other than those within the
forts, for use in repelling assaults and making counter-attacks.
 We will deal first with the stationary machine guns. It will not be
necessary to provide a machine gun for each cone mounting, as the
guns can be easily carried from one cone to another as required.
Shields should be used with all stationary mountings, but must be
detachable, and should not be placed in position until required, as
they indicate the situation of the gun and are easily destroyed by
artillery.
The positions for machine-gun mountings in a fort must depend
on the size and construction of the work, the nature of the outer
defences, and particularly on the supporting works in the immediate
vicinity. Positions commanding the glacis and the entanglements, on
salients of works and enfilading ditches, and any dead ground where
the enemy may effect a lodgment, are suitable. Counterscarp
galleries at the angles of works flanking the ditch should invariably
have machine-gun positions, with a special loophole, long and
shallow, to enable them to sweep a wide area with fire.
The selection of the positions for machine guns in permanent
works belongs to the art of the engineer, and there is little that can
be said of their tactical employment. Fire should be reserved until
the attack has reached close range, and then only opened when the
target is large and vulnerable. Guns must be concealed by every
possible device and all the loopholes must be blinded. By night all
guns should be mounted and trained to sweep ground by which the
enemy must approach; when search-lights are not in use, the
elevation should be checked by the use of a clinometer and the
amount of traverse may be shown by chalk lines on the parapet, or
white stones placed in two rows. In this way accurate fire may be
brought to bear on the assault on the darkest night, and many night
attacks were repulsed with machine-gun fire by the Russians at Port
Arthur. Sir G. S. Clarke says: “The front faces of the forts were
retrenched in some cases by obstacles and a line of field parapet
across the terre-plein. These, with the assistance of machine guns
brought up at the last moment, enabled assaults of the breaches
formed by the mines to be repulsed.” Again, “The Russians used
machine guns with effect, frequently concealing them in light
blindages, so that their positions could not be detected until they
were brought to bear upon an attacking force.”
The war correspondent Mr. F. Villiers, in his book Three Months
with the Besiegers, speaking of the storming of West Panlung
Redoubt,[40] says: “The death-dealing machine guns of the Russians
in the casemates of the fort are playing ghastly havoc—such havoc
that only a score or more of Ouchi’s battalions reached the first ditch
of the fence, where they threw themselves panting into the grateful
cover of the pits their own artillery have torn.”
The number of machine guns allotted to the permanent works of
Port Arthur is given as 38 by the United States Official Report, while
Nojine, in The Truth about Port Arthur, gives them in detail as 28,
the distribution of which is shown in red figures on the map at the
end of Chapter IX. The ten guns unaccounted for were probably
mobile, and used for the defence of the harbour and the various
landing-places in the neighbourhood of the fortress.
The mobile machine guns of the fortress will be used on the
advanced line of defence with the mobile troops, and should be
divided into two—those allotted to the outposts and those allotted to
the local reserve.[41] Those allotted to the outposts must be placed
in carefully selected positions commanding the approaches to the
section of the defence to which they have been posted. These
positions will usually be in minor works such as redoubts,
emplacements, and lunettes, and they will be selected for their good
field of fire, particular attention being paid to their command of dead
ground in front of other works. Great care and trouble must be taken
in concealing the guns and providing them with good cover, not only
from rifle fire, but also from artillery. An endeavour should be made
to command all wire entanglements along the front with machine
guns, and the angles of traverse of each gun should be carefully laid
off and marked in white paint or tape, so that they may be used in
the dark accurately to sweep their area of ground. Too much stress
cannot be laid on the importance of the accurate laying and sighting
of machine guns by day for use at night, and it must be
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com