Hands-on learning Deep learning
中文 | English
1 Introduction
This project is used to realize and read the code of "Hands-on Deep Learning", which is mainly in the direction of NLP at present.
Most of the book features executable code because we believe in the importance of interactive learning experiences in deep learning. At present, a
These intuitions can only be developed by trial and error, tweaking the code slightly, and observing the results. Ideally, an elegant mathematical theory would be accurate
Show us how to tune the code to achieve the desired result. Unfortunately, such an elegant theory is not yet available. Despite our best efforts
Force, but still lack a formal explanation of the various techniques, both because the mathematics of describing these models can be very difficult, and because of these masters
Serious study of the problem has only recently come to a head. We hope that with the development of deep learning theory, future editions of this book will be able to be released in the current edition without
The law provides a place for insight.
Sometimes, to avoid unnecessary duplication, we wrap functions, classes, and so on that are frequently imported and referenced throughout the book in the d2l package. For items to be saved to
Any block of code in the package, such as a function, a class, or multiple imports, is marked with #@save. We provide this in Section 16.6
These functions and classes are described in detail.
2 Update
1 updated the code and text for chapter 15
The directories are:
-15.1 Sentiment Analysis and data sets
-15.2 Sentiment analysis: Using recurrent neural networks
-15.3 Sentiment analysis: Use convolutional neural network
-15.4 Natural Language Inference and data sets
-15.5 Natural Language Inference: Using attention
-15.6 Fine-tune BERT for sequence-level and lexical level applications
-15.7 Natural Language Inference: Fine-tuning BERT
3 Hands-on Learning Deep Learning
The source and content website: https://zh-v2.d2l.ai/chapter_preliminaries/ndarray.html