Here are of deep learning javascript series directory, the book from the livebook.manning.com/book/deep-l… I will load the Chinese translation periodically. Angels who are interested in Machine Learning with javascript can pay attention to me. Let’s study and discuss together

1. Introduction:Deep learning and javascript

The contents include:

  • What is deep learning and its relation to artificial intelligence (AI) and machine learning
  • What makes deep learning stand out among the various machine learning technologies, and what has led to the current “deep learning revolution”
  • Reasons to use tensorflow.js for deep learning in JavaScript
  • The overall content of the book

2. The Beginning:Linear regression in tensorflow.js

The contents include:

  • Basic neural network example for simple machine learning tasks for linear regression
  • Tensors and tensor operations
  • Basic neural network optimization

3.Nonlinear regression and weighting

The contents include:

  • What is nonlinearity and how does nonlinearity in the hidden layers of neural networks increase the capacity of the network and lead to better prediction accuracy
  • What are hyperparameters and their adjustment methods
  • Phishing site detection examples to introduce binary classification through nonlinearity at the output layer
  • An example of iris will introduce multi-category classification and its difference from binary classification

4.Convolutional neural networks recognize images and sounds

The contents include:

  • How can images and other perceptual data, such as audio, be represented as multidimensional tensors
  • What are convolutional neural networks (ConvNets), how do they work and why are they particularly suited to machine learning tasks for images
  • How to write and train a convolutional network in tensorflow.js to solve the task of sorting numbers
  • How to train models in Node.js for faster training
  • How to use convolutional networks on audio data for speech recognition

5.Transfer learning: Reusing pre-trained neural networks

The contents include:

  • What is transfer learning? Why is it better than a beginner training model for multiple types of problems?
  • How can I leverage the feature extraction capabilities of the latest pre-trained convolutional neural networks (ConvNets) by converting from Keras and importing it into Tensorflow.js
  • What SymbolicTensor is and how they can help you achieve flexible “plug and play” for model components
  • Why should you freeze other layers during transfer learning to update only certain layers of the model
  • How to replace the output layer of the pre-trained convolutional network with a new output layer to solve different types of transfer learning tasks and data sets
  • What is fine-tuning and how can it help you get more accurate models from transfer learning
  • How to use transfer learning to implement object detection in tensorflow.js

6.The data set

The contents include:

  • How do I use the TF.data API to model training large data sets
  • Explore data to identify and resolve potential problems
  • How can data extensions be used to create new “pseudo-examples” to improve model quality

7.Visualize data and models

The contents include:

  • How do I perform custom visualization of data using TFJS-VIS
  • After training the model and getting results, how to peep into the inner working principle of the model

8.Generic workflows for underfitting, overfitting and machine learning

The contents include:

  • Why is it important to visualize the model training process, and what are the important things to look for
  • How to visualize and understand underfitting and overfitting
  • The main way to deal with overfitting: regularization, and how to visualize its effects
  • What is the general workflow for machine learning, what steps are involved, and why is it an important secret to guide all supervised machine learning tasks

9.Deep learning sequences and texts

The contents include:

  • How sorted data differs from unsorted data
  • Which deep learning techniques are appropriate for problems involving sequential data
  • How to represent text data in deep learning, including one-key encoding, multi-key encoding and word embedding
  • What are recursive neural networks (RNN) and why are they suitable for sequential problems
  • What is one-dimensional convolution and why is it an alternative to RNN
  • The unique properties of sequence-to-sequence (SEQ2SEQ) tasks and how can they be solved using attention mechanisms

10.Generative deep learning

The contents include:

  • What is generative deep learning, its applications, and how does it differ from the deep learning tasks we’ve seen so far
  • How to generate text using recursive neural networks
  • With the example of VAE, what are the potential Spaces and how they are formed to generate novel images
  • The foundation of generative adversarial networks (GAN)

11.Basis of reinforcement learning

The contents include:

  • How reinforcement learning (RL) differs from the supervised learning described in previous chapters
  • Basic paradigms of reinforcement learning: Agency, environment, behavior, and reward, and their interactions
  • The general ideas behind the two main approaches to solving RL problems: the policy-based approach and the value-based approach
  • Example based policy-based RL algorithm: Solving the shopping cart problem using the policy gradient (PG) approach
  • Through the RL algorithm based on Q value as an example: using deep Q network (DQN) to solve the snake game.

12.Test, optimize, deploy the model

The contents include:

  • Importance and practical guidelines for testing and monitoring machine learning code
  • How can models be trained and optimized in tensorflow.js or converted to tensorflow.js for faster loading and inference
  • How do YOU deploy the Tensorflow.js model to a variety of platforms and environments, from browsers to mobile applications, from desktop applications to single board computers

13.Summary and Prospect

The contents include:

  • Review advanced concepts and ideas about artificial intelligence and deep learning
  • This book provides a brief introduction to various deep learning algorithms, when they are useful, and how to implement them in tensorflow.js.
  • Pre-training model from the Tensorflow.js ecosystem
  • Current limitations of deep learning; An educated forecast of deep learning trends for the next few years
  • The latest guide on how to further improve your deep learning knowledge and keep up with The Times.