The 100 Page Machine Learning Book
This is a short summary of my notes from this book
Table of Contents
- The 100 Page Machine Learning Book
- Book Reading Details
- Chapter 0 : Preface
- Chapter 1 : Introduction
- Chapter 2 : Notation and Definitions
- Chapter 3 : Fundamental Algorithms
- Chapter 4 : Anatomy of a Learning Algorithm
- Chapter 5 : Best Practice
- Chapter 6 : Neural Networks and Deep Learning
- Chapter 7 : Problems and Solutions
- Chapter 8 : Advanced Practice
- Chapter 9 : Unsupervised Learning
- Chapter 10: Other Forms of Learning
- Chapter 11: Conclusion
Book Reading Details
| Tag | Information | Tag | Information |
|---|---|---|---|
| Book Name | The 100 Page Machine Learning Book | Tagline | |
| Author | Andriy Burkov | Year | 2019 |
| Pages | 136 | Chapters | 11 |
| Start Date | 25-Sep-2023 | End Date | |
| Status | In Progress | Total Days | 20 days and counting … |
Chapter 0 : Preface
- Machines don’t learn !!!
- They only come up with a mathematical function that maps a set of inputs to a set of outputs
- If input values are from a different distribution than what the machine was trained on, then it would not give the right output. This is why it is not like the learning that we humans do
- “Machine Learning” is only a marketing term coined by Arthur Samuel in 1959 while at IBM
- The term “Learning” in “Machine Learning” is only used as an analogy and not literally to mean the learning than we humans do
Chapter 1 : Introduction
-
Types of Learning
-
Supervised Learning
1.1 Creates a model using labelled examples ( or target is known )
1.2 Commonly used for Predictions and Classification problems
-
Semi-supervised Learning
2.1 Contains a large number of unlabelled examples and a small number of labelled examples
2.2 Saves costs for labelling a large dataset.
-
Unsupervised Learning
3.1 Creates a model using unlabelled examples
3.2 Converts input data into another vector or a value that is later on used to solve a practical problem
-
Reinforcement Learning
4.1 Creates a policy ( a function similar to a model ) that outputs an action to execute for the give “state” the machine is in
4.2 This is not covered in this book
-
-
Support Vector Machine (SVM) algorithm is introduced to explain how Supervised Learning works
Chapter 2 : Notation and Definitions
- Lots of mathematical concepts are defined and their intuition is explained
- Matrix
- Vector Operations
- Functions
- Derivative
- Random Variable
- Parameters
- Hyperparameters
- Classification algorithms predict the target class ( label )
- Regression algorithms predict a real-valued label
- Model-Base Learning is about learning the optimum “parameters” for a model
- Instance-Based Learning is about using the entire dataset as a model. Ex: KNN
- Shallow Learning is where the model parameters are learnt directly from the features of the training example
- Deep Learning is where most of the model parameters are learnt from the outputs of the preceding layers ( as in a deep neural network )