Название: The Little Learner: A Straight Line to Deep Learning
Автор: Daniel P. Friedman, Anurag Mendhekar
Издательство: The MIT Press
Год: 2023
Страниц: 440
Язык: английский
Формат: epub (true)
Размер: 10.2 MB

A highly accessible, step-by-step introduction to Deep Learning, written in an engaging, question-and-answer style.

The Little Learner introduces Deep Learning from the bottom up, inviting students to learn by doing. With the characteristic humor and Socratic approach of classroom favorites The Little Schemer and The Little Typer, this kindred text explains the workings of deep neural networks (DNN) by constructing them incrementally from first principles using little programs that build on one another. Starting from scratch, the reader is led through a complete implementation of a substantial application: a recognizer for noisy Morse code signals. Example-driven and highly accessible, The Little Learner covers all of the concepts necessary to develop an intuitive understanding of the workings of deep neural networks, including tensors, extended operators, gradient descent algorithms, artificial neurons, dense networks, convolutional networks, residual networks, and automatic differentiation.

Deep Learning, an emerging area of Artificial Intelligence, has revolutionized the way problems are solved, be it winning at Go, recognizing cats in pictures, or asking a smart speaker to order pizza. The most beautiful thing about Deep Learning is how simple pieces come together to solve large, complex problems. How can we understand what makes these Deep Learning tools work? Our approach is to build them, a little bit at a time, and watch them work.

We package our ideas as little Scheme programs. Scheme allows our thoughts to be expressed clearly and directly, and with minimal fuss. It is a language that assumes very little and gets out of the way quickly, so that the code speaks for itself. We use a very small subset of Scheme: define (or let) allows a global (or local) name to be given to a value, lambda creates a function as a value, and cond dispatches over a sequence of (test value) pairs.

We have collected the functions and syntactic extensions necessary for the code in this book into a MAchine Learning Toolkit package, called Malt. Malt is a package in Racket, which is a superset of our small subset of Scheme. The package includes our code and examples as well as the tools necessary to experiment with them.

• Conversational style, illustrations, and question-and-answer format make Deep Learning accessible and fun
• Incremental approach constructs advanced concepts from first principles
• Presents key ideas of Machine Learning using a small, manageable subset of the Scheme language
• Suitable for anyone with knowledge of high school math and some programming experience