regression2chatgpt

Deconstructing Large Language Models: From Linear Regression to General Artificial Intelligence

This book is currently in the editing process and will be available soon.

Description

For classic models in AI(artificial intelligence), the tools, such as PyTorch, have provided well-encapsulated implementations, making their usage relatively straightforward. However, due to engineering considerations, these implementations introduce excessive details into the code, complicating model understanding. This book aims to enhance reader comprehension by re-implementing core model parts with detailed annotations. While explaining complex algorithms in human language can be challenging, reading the code proves more intuitive.

The code relies on external libraries, with installation commands given at the beginning of the scripts. Rerunning might yield slight result variations due to random numbers, but the overall impact is minimal. For large language models, running the code on a GPU is crucial to avoid a significant increase in computation time.

Outline

• ch03_linear: Linear Regression
• ch04_logit: Logistic Regression
• ch05_econometrics: Insights from Econometrics
• ch06_optimizer: Optimization Algorithms
• ch07_autograd: Backpropagation
• ch08_mlp: Multilayer perceptron
• ch09_cnn: Convolutional Neural Network
• ch10_rnn: Recurrent Neural Network
• ch11_llm: Large Languague Model
• ch12_rl: Reinforcement Learning
• ch13_others: Other Classic Models