Hands-on Deep Learning with TensorFlow

Women GDG
Techmakers Ahmedabad
Interest
Google NGRAM & Google Trends

Trend of “Deep Learning” in Google Web Searches

Hype or Reality?
Quotes

I have worked all my life in Machine Learning, and I have never seen
one algorithm knock over benchmarks like Deep Learning
– Andrew Ng (Stanford & Baidu)

Deep Learning is an algorithm which has no theoretical

limitations of what it can learn; the more data you give and the
more computational time you provide, the better it is – Geoffrey
Hinton (Google)

Human-level artificial intelligence has the potential to help

humanity thrive more than any invention that has come before it –
Dileep George (Co-Founder Vicarious)

For a very long time it will be a complementary tool that human

scientists and human experts can use to help them with the
things that humans are not naturally good – Demis Hassabis (Co-Founder
DeepMind)
Hype or Reality?
Deep Learning at Google
Hype or Reality?
Deep Learning at Google
What is Learning?
A closer look at how Humans learn
Basic Terminologies

• Features
• Labels
• Examples
• Labelled example
• Unlabelled example
• Models (Train and Test)
• Classification model
• Regression model
What is Artificial
Intelligence?
Machine Learning -
Basics
Introduction

Machine Learning is a type of Artificial Intelligence that

provides computers with the ability to learn without being
explicitly programmed.

Machine Learning
Algorithm
Labeled Data

Training
Prediction

Learned Model Prediction

Data

Provides
various
technique
Machine Learning -
Basics
Learning Approaches

Supervised Learning: Learning with a labeled training set

Example: email spam detector with training set of already labeled emails

Unsupervised Learning: Discovering patterns in unlabeled data

Example: cluster similar documents based on the text content

Reinforcement Learning: learning based on feedback or reward

Example: learn to play chess by winning or losing
Machine Learning -
Basics
Problem Types

Classification Regression
(supervised – predictive) (supervised – predictive)

Clustering Anomaly Detection

(unsupervised – descriptive) (unsupervised– descriptive)
What is Deep
Learning?

Part of the machine learning field of learning representations

of data. Exceptional effective at learning patterns.

Utilizes learning algorithms that derive meaning out of data by

using a hierarchy of multiple layers that mimic the neural networks
of our brain.

If you provide the system tons of information, it begins to

understand it and respond in useful ways.
Feature Engineering

• A machine learning model can't directly see, hear, or sense input

examples. Machine learning models typically expect examples to be
represented as numbered vectors
• Feature engineering means transforming raw data into a feature
vector of 1’s and 0’s which Machine can understand
How does DL works?
Why DL over traditional ML?

• Machine Learning is set of algorithms that parse data, learn from

them, and then apply what they’ve learned to make intelligent
decisions
• The thing about traditional Machine Learning algorithms is that as complex
as they may seem, they’re still machine like
• They need lot of domain expertise, human intervention only capable of
what they’re designed for; nothing more, nothing less
• For AI designers and the rest of the world, that’s where deep learning
holds a bit more promise
Why DL over traditional ML?

• Deep Learning requires high-end machines contrary to traditional Machine

Learning algorithms
• Thanks GPUs and TPUs
• No more Feature Engineering!!
• ML: most of the applied features need to be identified by an domain expert
in order to reduce the complexity of the data and make patterns more
visible to learning algorithms to work
• DL: they try to learn high-level features from data in an incremental
manner.
Why DL over traditional ML?

• The problem solving approach:

• Deep Learning techniques tend to solve the problem end to end
• Machine learning techniques need the problem statements to break
down to different parts to be solved first and then their results to be
combine at final stage
• For example for a multiple object detection problem, Deep
Learning techniques like Yolo net take the image as input and provide the
location and name of objects at output
• But in usual Machine Learning algorithms uses SVM, a bounding box
object detection algorithm then HOG as input to the learning algorithm in
order to recognize relevant objects
What changed?
Old wine in new
bottles

Big Data Computation Algorithmic

(Digitalization (Moore’s Law, Progress
) GPUs)
When to use DL or not over Others?

1. Deep Learning outperforms other techniques if the data size is large. But

with small data size, traditional Machine Learning algorithms are
preferable
2. Finding large amount of “Good” data is always a painful task but hopefully
not now on, Thanks to the all new Google Dataset Search Engine 
3. Deep Learning techniques need to have high end infrastructure to train
in reasonable time
4. When there is lack of domain understanding for feature introspection,
Deep Learning techniques outshines others as you have to worry less
about feature engineering
5. Model Training time: a Deep Learning algorithm may take weeks or
months where as, traditional Machine Learning algorithms take few
seconds or hours
6. Model Testing time: DL takes much lesser time as compare to ML
7. DL never reveals the “how and why” behind the output- it’s a Black Box
8. Deep Learning really shines when it comes to complex problems such
as image classification, natural language processing, and speech
recognition
9. Excels in tasks where the basic unit (pixel, word) has very little meaning in
itself, but the combination of such units has a useful meaning
Got interested??
Applications of Deep Learning
Try it yourself: https://www.clarifai.com/demo
Applications
Deep Learning
Try it yourself: https://www.paralleldots.com/sentiment-analysis
Applications
Deep Learning
Applications
Deep Learning
Applications
Deep Learning
The Big Players
Indian Startups
TensorFlow

● TensorFlow is an open-source library for Machine Intelligence

● It was developed by the Google Brain and released in 2015
● It provides high-level APIs to help implement many machine learning
algorithms and develop complex models in a simpler manner.

● What is a tensor?
● A mathematical object, analogous to but more general than a vector,
represented by an array of components that are functions of the
coordinates of a space.
● TensorFlow computations are expressed as stateful dataflow graphs.
● The name TensorFlow derives from the operations that such neural
networks perform on multidimensional data arrays known as
‘tensors’.
Why TensorFlow?

This is a dialogue between 2 persons on “Why Tensorflow?”

• Person 1: Well it’s an ML framework!!

• Person 2: But isn’t it is a complex one, I know a few which are very simple
and easy to use like Sci-Kit learn, PyTorch, Keras, etc. Why to use
Tensoflow?
• Person 1: Ok, Can you implement your own Model in Sci-Kit learn and scale
it if you want?
• Person 2: No. Ok but then for Deep Learning, why not to use Keras or
PyTorch? It has so many models already available in it.
• Person 1: Tensorflow is not only limited to implementing your own models. It
also has lot many models available in it. And apart from that you can do a
large scale distributed model training without writing complex infrastructure
around your code or develop models which need to be deployed on mobile
platforms.
• Person 2: Ok. Now I understand “Why Tensorflow?”
What TensorFlow does for You?

• Calculation on large data set includes so much of boring Mathematics,

Equations, etc.
• Implementation of it from scratch all the time is a bit cumbersome task
• Code will also be complex
• TensorFlow is Scalable

• What is Scalability?
• Think of a Smart Traffic Management System of Ahmedabad city :D
• Roads having 4 lanes, number of cross-roads, and not for a single area
-> so much of Computation Data
• Streaming data, Continues data, Decision in real-time
• Single Computer cannot handle it
• Solution?
• Assign computers area/range wise then integrate all of them, this is no
more a complex task today
• Why?
• TensorFlow will take care of it!!
• It can scale the hardware/software requirements by clustering as per the
requirement
What TensorFlow does for You?

• Creates own environment, takes care of everything you will need!

• Manage memory allocations
• Create some variable, you can scale it, make them global
• Statistical and Deep Learning both methods can be implemented
• 3D list, computation of Graph is fast because of the very powerful and
Optimised Data Structure
• Good for Research and Testing
• Useful for Production level coding
How do you Classify these Points?
Okay, how do you Classify these Points?
Okay okay, but now?

Non linearities are tough to model.

In complex datasets, the task
becomes very cumbersome. What
is the solution?
Inspired by the human Brain

An artificial neuron contains a nonlinear activation function and has

several incoming and outgoing weighted connections.

Neurons are trained to filter and detect specific features or

patterns (e.g. edge, nose) by receiving weighted input,
transforming it with the activation function und passing it to the
outgoing connections.
Modelling a Linear Equation
How to deal with Non-linear Problems?

We added a hidden layer of intermediary values. Each yellow node in the

hidden layer is a weighted sum of the blue input node values. The output is
a weighted sum of the yellow nodes.
Is it linear? What are we missing?
Activation Functions

Non-linearity is needed to learn complex (non-linear)

representations of data, otherwise the NN would be
just a linear function.
Non-linear Activation Functions
Gradient Descent

Gradient Descent finds the (local) minimum of the cost function (used to
calculate the output error) and is used to adjust the weights
Gradient Descent

• Convex problems have only one minimum; that is, only one place where the
slope is exactly 0. That minimum is where the loss function converges
• The gradient descent algorithm then calculates the gradient of the loss curve
at the starting point. In brief, a gradient is a vector of partial derivatives
• A gradient is a vector and hence has magnitude and direction
• The gradient always points in the direction of the minimum. The gradient
descent algorithm takes a step in the direction of the negative gradient in
order to reduce loss as quickly as possible
Gradient Descent

• The algorithm given below signifies Gradient

Descent algorithm
• In our case,
• Өj will be wi
• 𝞪 is the learning rate
• J(Ө) is the cost function
The Learning Rate

• Gradient descent algorithms multiply the gradient by a scalar known as the

learning rate (also sometimes called step size) to determine the next point.
• For example, if the gradient magnitude is 2.5 and the learning rate is 0.01,
then the gradient descent algorithm will pick the next point 0.025 away from
the previous point.
• A Hyperparameter!
• Think of it as in real life. Some of us slow learners while some others are
quicker learners
• Small learning rate -> learning will take too long
• Large learning rate -> may overshoot the minima
But how the model will LEARN?

BACKPROPAGATION
Deep Learning
The Training Process

Forward it trough
the network to
Sample labeled data get predictions

Update the Backpropagat

connection weights e the errors

Learns by generating an error signal that measures the difference between the
predictions of the network and the desired values and then using this error
signal to change the weights (or parameters) so that predictions get more
accurate.
Still not so Perfect!
Backprop can go wrong

• Vanishing Gradients:
• The gradients for the lower layers (closer to the input) can become very
small. In deep networks, computing these gradients can involve taking the
product of many small terms

• Exploding Gradients:
• If the weights in a network are very large, then the gradients for the lower
layers products of many large terms. In this case you can have exploding
gradients: gradients that get too large to converge
Ooooooverfitting = Game Over

• An overfit model gets a low loss during training but does a poor job predicting
new data
• Overfitting is caused by making a model more complex than necessary.
• The fundamental tension of machine learning is between fitting our data well,
but also fitting the data as simply as possible
Solution
Dropout Regularization

It works by randomly "dropping out" unit activations in a network for a single

gradient step. The more you drop out, the stronger the regularization:

0.0 -> No dropout regularization.

1.0 -> Drop out everything. The model learns nothing
values between 0.0 and 1.0 -> More useful
Softmax

Now the problem with sigmoid function in multi-class classification is that the
values calculated on each of the output nodes may not necessarily sum up to
one.

The softmax function used for multi-classification model returns the probabilities
of each class.
Game Time!!

Visit kahoot.it
Game PIN: 508274
Convolutional Neural Nets (CNN)

Convolution layer is a feature detector that automagically learns to filter out not
needed information from an input by using convolution kernel.

Pooling layers compute the max or average value of a particular feature over a
region of the input data (downsizing of input images). Also helps to detect
objects in some unusual places and reduces memory size.
Convolution…! ;)
Convolution
Max Pooling
Let’s build our first CNN

Visit: https://
colab.research.google.com/drive/1arAJnnTn0wI3KoSSJHg_Hjw
40VPPMtP0
Takeaways

Humans are genius!!!

Machines that learn to represent the world from experience.

Deep Learning is no magic! Just statistics in a black box, but

exceptional effective at learning patterns.

We haven’t figured out creativity and human-empathy.

Neural Networks are not the solution to every problem.

Transitioning from research to consumer products. Will make the

tools you use every day work better, faster and smarter.
This was just a Start!
Online Courses

• DL Specialization: https://www.deeplearning.ai/
• Deep Learning A-Z™: Hands-On Artificial Neural Networks:
https://www.udemy.com/deeplearning/?
siteID=AKW.sgcfqI8oqN6eoMfxusNIligTml0Iw&LSNPUBID=AKW*sgcfqI8
• The Canonical Machine Learning Course:
https://www.coursera.org/learn/machine-learning
• CMU ML Course: http://www.cs.cmu.edu/~tom/10701_sp11/lectures.shtml
• University of Washington: https://www.coursera.org/learn/ml-foundations
This was just a Start!
Visit Blogs

• Machine Learning Google News

• Machine Learning Mastery
• Medium: https://medium.com/towards-data-science/machine-learning/home
• Andej Karpathy: http://karpathy.github.io/
• I am Trask: http://iamtrask.github.io/
• Quora: https://www.quora.com/topic/Machine-Learning
• AI StackExchange: https://ai.stackexchange.com/
This was just a Start!
Useful Links (Extremely!)

• Over 200 of the Best Machine Learning, NLP, and Python Tutorials:  https
://medium.com/machine-learning-in-practice/over-200-of-the-best-machine-le
arning-nlp-and-python-tutorials-2018-edition-dd8cf53cb7dc

• Awesome Deep Learning:

https://github.com/ChristosChristofidis/awesome-deep-learning

• Machine Learning Glossary:

https://developers.google.com/machine-learning/glossary/
References

• https://
towardsdatascience.com/why-deep-learning-is-needed-over-traditional-mac
hine-learning-1b6a99177063
• https://iamtrask.github.io/2015/07/12/basic-python-network/
• https://www.youtube.com/watch?v=BmkA1ZsG2P4
• https://
www.slideshare.net/LuMa921/deep-learning-a-visual-introduction?from_acti
on=save
• https://developers.google.com/machine-learning/crash-course/
Questions…??
Comments
Suggestions 
 Happy Learning!

Charmi Chokshi
AI and Data Enthusiast
Final year ICT Engineering
Student at Ahmedabad
University

Let’s Connect!
• LinkedIn
• Github

Women GDG
Techmakers Ahmedabad