Prologue
Icons
CHAPTER 1 Introduction:
Learning Is Key
A Little Bit Of History
Key Methodologies Covered In This Book
Classical Mathematical Modelling
Machine Learning Is Different
Simplicity Leading To Complexity
CHAPTER 2 General Issues:
Jargon And Notation
Scaling
Measuring Distances
Curse Of Dimensionality
Principal Components Analysis
Maximum Likelihood Estimation
Confusion Matrix
Cost Functions
Gradient Descent
Training, Testing And Validation
Bias And Variance
Lagrange Multipliers
Multiple Classes
Information Theory And Entropy
Natural Language Processing
Bayes Theorem
CHAPTER 3 K Nearest Neighbours:
Executive Summary
What Is It Used For?
The Algorithm
Problems With KNN
Example: Heights and weights
Regression
CHAPTER 4 K Means Clustering:
Executive Summary
What Is It Used For?
What Does K Means Clustering Do?
Scree Plots
Example: Crime in England, a 13-dimensional example
Example: Volatility
Example: Interest rates and inflation
Example: Interest rates, inflation and GDP growth
CHAPTER 5 Naive Bayes Classifier:
Executive Summary
What Is It Used For?
Using Bayes Theorem
Application Of NBC
In Symbols
Example: Political speeches
CHAPTER 6 Regression Methods:
Executive Summary
What Is It Used For?
Linear Regression In Many Dimensions
Logistic Regression
Example: Political speeches again
Other Regression Methods
CHAPTER 7 Support Vector Machines:
Executive Summary
What Is It Used For?
Hard Margins
Example: Irises
Lagrange Multiplier Version
Soft Margins
Kernel Trick
CHAPTER 8 Self-Organizing Maps:
Executive Summary
What Is It Used For?
The Method
The Learning Algorithm
Example: Grouping shares
Example: Voting in the House of Commons
CHAPTER 9 Decision Trees:
Executive Summary
What Is It Used For?
Example: Magazine subscription
Entropy
Overfitting And Stopping Rules
Pruning
Numerical Features
Regression
Looking Ahead
Bagging And Random Forests
CHAPTER 10 Neural Networks:
Executive Summary
What Is It Used For?
A Very Simple Network
Universal Approximation Theorem
An Even Simpler Network
The Mathematical Manipulations In Detail
Common Activation Functions
The Goal
Example: Approximating a function
Cost Function
Backpropagation
Example: Character recognition
Training And Testing
More Architectures
Deep Learning
CHAPTER 11 Reinforcement Learning:
Executive Summary
What Is It Used For?
Going Offroad In Your Lamborghini 400 GT
A First Look At Blackjack
The Classical MDP Approach In Noughts & Crosses
Example: The multi-armed bandit
Getting More Sophisticated 1: Known environment
Example: A maze
Value Notation
The Bellman Equations
Optimal Policy
The Role Of Probability
Getting More Sophisticated 2: Model free
Monte Carlo Policy Evaluation
Temporal Difference Learning
Pros And Cons: MC v TD
Finding The Optimal Policy
Sarsa
Q Learning
Example: Blackjack
Large State Spaces
Datasets
Epilogue
Index
