Neural Networks From Scratch
By Harrison Kinsley, Daniel Kukieła
Language: English
Category: Ebook - Computer/Internet
Categories: Computer/Internet
Uploader: ParSal
Uploaded At: 2021-08-27 05:37:13
Published 2020
Description
"Neural Networks From Scratch" is a book intended to teach you how to build neural networks on your own, without any libraries, so you can better understand deep learning and how all of the elements work. This is so you can go out and do new/novel things with deep learning as well as to become more successful with even more basic models.
This book is to accompany the usual free tutorial videos and sample code from youtube.com/sentdex. This topic is one that warrants multiple mediums and sittings. Having something like a hard copy that you can make notes in, or access without your computer/offline is extremely helpful. All of this plus the ability for backers to highlight and post comments directly in the text should make learning the subject matter even easier.
First off, there's none of that "intro to programming" padding of any kind! The book starts off with a brief outline of what neural networks are and some general background on the structure of machine learning algorithms, as I expect some people will have neural networks as their first machine learning algorithm and be a bit confused about terms like "features" and "labels" ...etc.
Within short order, we're coding our first neurons, creating layers of neurons, building activation functions, calculating loss, and doing backpropagation with various optimizers. Everything is covered to code, train, and use a neural network from scratch in Python.
Everything we do is shown first in pure, raw, Python (no 3rd party libraries). Then you're shown how to use NumPy (the go-to 3rd party library in Python for doing mathematics) to do the same thing, since learning more about using NumPy can be a great side-benefit of the book.
In this book, you will:
- Code a neuron
- Learn how to connect these neurons in layers
- Program activation functions: Rectified Linear (ReLU), Softmax, Sigmoid, and Linear
- Calculate cross-entropy loss
- Code and perform gradient computations using backpropagation and parameter updates using optimizers: Stochastic Gradient Descent (SGD), AdaGrad, RMSprop, and Adam
- ...and most importantly: build and train a fully working neural network, from scratch, in Python...learning a ton along the way!
Certain concepts, while also explained by text and images, can also be supplemented with animations. In the book, there will be QR codes to help explain certain concepts, for example (you need a QR-scanning app on your phone, and feel free to give this one a go):