Machine learning involves using algorithms to learn from data and improve performance on tasks without being explicitly programmed. The document discusses what machine learning is, why it is used, examples of applications, different types of machine learning systems, and main challenges including having insufficient or unrepresentative training data, poor quality data, irrelevant features, and overfitting or underfitting models.
Machine learning involves using algorithms to learn from data and improve performance on tasks without being explicitly programmed. The document discusses what machine learning is, why it is used, examples of applications, different types of machine learning systems, and main challenges including having insufficient or unrepresentative training data, poor quality data, irrelevant features, and overfitting or underfitting models.
Machine learning involves using algorithms to learn from data and improve performance on tasks without being explicitly programmed. The document discusses what machine learning is, why it is used, examples of applications, different types of machine learning systems, and main challenges including having insufficient or unrepresentative training data, poor quality data, irrelevant features, and overfitting or underfitting models.
Machine learning involves using algorithms to learn from data and improve performance on tasks without being explicitly programmed. The document discusses what machine learning is, why it is used, examples of applications, different types of machine learning systems, and main challenges including having insufficient or unrepresentative training data, poor quality data, irrelevant features, and overfitting or underfitting models.
Download as PPTX, PDF, TXT or read online from Scribd
Download as pptx, pdf, or txt
You are on page 1/ 49
Introduction to Machine
Learning By Jawad Haider Outline:
• What is ML? • Why Use ML? • Examples of Application • Types of Machine Learning Systems • Main Challenges of ML * • What is ML? The science ( and Art) of Programming computers so they can learn from Data.
• An Engineering oriented notion:
A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with Experience E. What is ML? • Traditional Programming Approach - A problem - Rules to solve the problem ( get some result)
Input Output result
Rules
• ML Approach (will be clear next)
Input + Output Results Figure out the
Rules Why Use ML?
• Consider how you would write a spam
filter using traditional programming? Instead use Machine Learning
• To study and solve problems that are
either too complex for traditional approaches or have no known algorithm. • ML Can help humans learn. Summary: • Problems for which existing solutions require a lot of fine-tuning or long lists of rules: one Machine Learning algorithm can often simplify code and perform better than the traditional approach. • Complex problems for which using a traditional approach yields no good solution: the best Machine Learning techniques can perhaps find a solution. • Fluctuating environments: a Machine Learning system can adapt to new data. • Getting insights about complex problems and large amounts of data. Question:
• Where not to Use ML?
Outline:
• What is ML? • Why Use ML? • Examples of Application • Types of Machine Learning Systems • Main Challenges of ML * Examples of Applications of ML? • Analyzing images of products on a product line to automatically classify them Examples of Applications of ML?
• Detecting tumors in Brain
Examples of Applications of ML? • Automatically classifying news article Examples of • Automatically flagging offensive comments on discussion forums Applications of ML? Embedding projector - visualization of high-dimensional data (tensorflo w.org) Examples of Applications of ML? • Summarizing long documents automatically (TextRank Algo) Examples of Applications of ML?
• Creating a chatbot or a personal
assistant Examples of Applications of ML?
• Forecasting your company’s revenue next
year, based on many performance metrics Examples of Applications of ML? • Analyzing images of products on a product line to automatically classify them • Detecting tumors in Brain • Automatically classifying new article • Automatically flagging offensive comments on discussion forums • Summarizing long documents automatically • Creating a chatbot or a personal assistant • Forecasting your company’s revenue next year, based on many performance metrics Examples of Applications of ML? Making your app react to voice commands Examples of Applications of ML? • Representing a complex, high- dimensional dataset in a clear and insightful diagram Examples of Applications of ML? Recommending a product that a client may be interested in, based on past purchases Examples of Applications of ML? Building an intelligent bot for a game Examples of Applications of ML? • Making your app react to voice commands • Representing a complex, high-dimensional dataset in a clear and insightful diagram • Recommending a product that a client may be interested in, based on past purchases • Building an intelligent bot for a game • And many many more… Outline:
• What is ML? • Why Use ML? • Examples of Application • Types of Machine Learning Systems • Main Challenges of ML * • Based on the criteria: 1. Whether or not the system is trained with
Types of 2. human supervision Whether or not the system can learn
ML systems 3. incrementally on the fly Whether they do simple comparison or detect patterns in the training data and build a predictive model (like scientists). Whether or not the system is trained with human supervision • Supervised Learning Example: K-Nearest Neighbors Linear Regression Logistic Regression Support Vector Machines Decision Trees and Random forests Neural Networks ** Training set, Testing Set Whether or not the system is trained with human supervision • Unsupervised Learning - Clustering - Anomaly Detection - Novelty Detection - Visualization - Dimensionality Reduction - Association Rule Learning Whether or not the system is trained with human supervision • Semi-supervised Learning
• Reinforcement learning AI Learns to Park - Deep Reinforcement Learning - YouTube • (6980) OpenAI Plays Hide and Seek…and Breaks The Game! 🤖 – YouTube • (6980) AlphaZero : DeepMind's New Chess AI | Two Minute Papers #216 – YouTube • Based on the criteria: 1. Whether or not the system is trained with
Types of 2. human supervision Whether or not the system can learn
ML systems 3. incrementally on the fly Whether they do simple comparison or they detect patterns in the training data and build a predictive model (like scientists). Whether or not the system can learn incrementally on the fly
• Batch and Online Learning
Batch Learning (offline learning): Incapable of learning incrementally;
Online Learning: you can train the system
incrementally by feeding it data instances sequentially, either individual or in small groups (mini-batches). • Based on the criteria: 1. Whether or not the system is trained with human supervision
Types of ML 2. Whether or not the system can learn
incrementally on the fly
systems I m skipping this one:
3. Whether they do simple comparison or they detect patterns in the training data and build a predictive model (like scientists). Outline:
• What is ML? • Why Use ML? • Examples of Application • Types of Machine Learning Systems • Main Challenges of ML * A short Quiz Main Challenges of ML (data Related)
1. Insufficient Quantity of Training
Data - The unreasonable Effectiveness of Data ( link)
“These results suggest that we may want to
consider the tradeoff between spending time and money on algorithm development versus spending it on corpus development” Main Challenges of ML (Data Related)
2. Nonrepresentative Training Data
Solution: Avoid Sampling bias (Coded bias)
3. Poor-Quality Data Solution: Data Preprocessing: Discard Outliers, fill in missing values, ignore attribute etc.
4. Irrelevant Features Solution: Feature engineering = Feature Selection + Feature Extraction Main Challenges of ML (Algorithm Related)
1. Overfitting & Underfitting the Training Data
• Machine Learning is about making machines get better at some task by learning from data, instead of having to explicitly code rules. • There are many different types of ML systems: supervised or not, batch or online, instance-based or model-based. • In an ML project you gather data in a training set, and you feed the training set to a learning algorithm. If the algorithm is model-based, it tunes some parameters to fit the model to the training set (i.e., to Stepping Back make good predictions on the training set itself), and then hopefully it will be able to make good predictions on new cases as well. If the algorithm is instance-based, it just learns the examples by heart and generalizes to new instances by using a similarity measure to compare them to the learned instances. • The system will not perform well if your training set is too small, or if the data is not representative, is noisy, or is polluted with irrelevant features (garbage in, garbage out). Lastly, your model needs to be neither too simple (in which case it will underfit) nor too complex (in which case it will overfit). Thank you References for further Readings
• 35179.pdf (googleusercontent.com) • Chapter 1: Hands-On Machine Learning with Scikit-Learn and TensorFlow (h ouchangtech.com) • Part 1: As Prerequisite: Deep Learning (deeplearningbook.org)
