This book provides an accessible introduction to the foundations of machine learning and deep learning in medicine for medical students, researchers, and professionals who are not necessarily initiated in advanced mathematics but yearn for a better understanding of this disruptive technology and its impact on medicine. Once an esoteric subject known to few outside of computer science and engineering departments, today artificial intelligence (AI) is a widely popular technology used by scholars from all across the academic universe. In particular, recent years have seen a great deal of interest in the AI subfields of machine learning and deep learning from researchers in medicine and life sciences, evidenced by the rapid growth in the number of articles published on the topic in peer-reviewed medical journals over the last decade. The demand for high-quality educational resources in this area has never been greater than it is today, and will only continue to grow at a rapid pace.
Expert authors remove the veil of unnecessary complexity that often surrounds machine learning and deep learning by employing a narrative style that emphasizes intuition in place of abstract mathematical formalisms, allowing them to strike a delicate balance between practicality and theoretical rigor in service of facilitating the reader’s learning experience. Topics covered in the book include: mathematical encoding of medical data, linear regression and classification, nonlinear feature engineering, deep learning, convolutional and recurrent neural networks, and reinforcement learning. Each chapter ends with a collection of exercises for readers to practice and test their knowledge.
This is an ideal introduction for medical students, professionals, and researchers interested in learning more about machine learning and deep learning. Readers who have taken at least one introductory mathematics course at the undergraduate-level (e.g., biostatistics or calculus) will be well-equipped to use this book without needing any additional prerequisites.
Author(s): Reza Borhani, Soheila Borhani, Aggelos K. Katsaggelos
Publisher: Springer
Year: 2022
Language: English
Pages: 200
City: Cham
Preface
Contents
1 Introduction
The Machine Learning Pipeline
Data Collection
Feature Design
Model Training
Model Testing
A Deeper Dive into the Machine Learning Pipeline
Revisiting Data Collection
Revisiting Feature Design
Revisiting Model Training
Revisiting Model Testing
The Machine Learning Taxonomy
Problems
References
2 Mathematical Encoding of Medical Data
Numerical Data
Categorical Data
Imaging Data
Time-Series Data
Text Data
Genomics Data
Problems
3 Elementary Functions and Operations
Different Representations of Mathematical Functions
Elementary Functions
Polynomial Functions
Reciprocal Functions
Trigonometric and Hyperbolic Functions
Exponential Functions
Logarithmic Functions
Step Functions
Elementary Operations
Basic Function Adjustments
Addition and Multiplication of Functions
Composition of Functions
Min–Max Operations
Constructing Complex Functions Using Elementary Functions and Operations
Problems
4 Linear Regression
Linear Regression with One-Dimensional Input
The Least Squares Cost Function
Linear Regression with Multi-Dimensional Input
Input Normalization
Regularization
Problems
Reference
5 Linear Classification
Linear Classification with One-Dimensional Input
The Logistic Function
The Cross-Entropy Cost Function
The Gradient Descent Algorithm
Linear Classification with Multi-Dimensional Input
Linear Classification with Multiple Classes
Problems
References
6 From Feature Engineering to Deep Learning
Feature Engineering for Nonlinear Regression
Feature Engineering for Nonlinear Classification
Feature Learning
Multi-Layer Neural Networks
Optimization of Neural Networks
Design of Neural Network Architectures
Problems
References
7 Convolutional and Recurrent Neural Networks
The Convolution Operation
Convolutional Neural Networks
Recurrence Relations
Recurrent Neural Networks
Problems
References
8 Reinforcement Learning
Reinforcement Learning Applications
Path-Finding AI
Automatic Control
Game-Playing AI
Autonomous Robotic Surgery
Automated Planning of Radiation Treatment
Fundamental Concepts
States, Actions, and Rewards in Gridworld
States, Actions, and Rewards in Cart–Pole
States, Actions, and Rewards in Chess
States, Actions, and Rewards in Radiotherapy Planning
Mathematical Notation
Bellman's Equation
The Basic Q-Learning Algorithm
The Testing Phase of Q-Learning
Tuning the Q-Learning Parameters
Q-Learning Enhancements
The Exploration–Exploitation Trade-Off
The Short-Term Long-Term Reward Trade-Off
Tackling Problems with Large State Spaces
Problems
References
Index
