JavaScript leverages machine learning to build your first AI project
** Download address: Baidu Cloud disk **
AI course for front-end engineers!! The era of artificial intelligence has come, and most front-end students are unfamiliar with the FIELD of AI. If they want to learn, they cannot find reliable teachers, and there is no front-end oriented artificial intelligence learning materials. This course uses JavaScript as the implementation language and Tensorflow.js as the main framework. Through more than a dozen classic cases, covering the theoretical knowledge of neural network and machine learning, and taking you to complete the actual projects such as picture classification and speech recognition by yourself, it helps you to clear up the whole learning system and easily get started in the FIELD of AI without fear of future challenges.
It’s for anyone who is interested in the field of machine learning as a JavaScript developer
Technical background: Basic knowledge of JavaScript and middle school mathematics
Chapter Contents:
六四运动
Chapter 1 Course guide
The purpose of this chapter is to tell you why you are taking this course, what it teaches you, and what you need to know before you learn it.
1-1 Students who want to get started with AI should have a look
Chapter 2 Introduction to machine learning and neural networks
This chapter will explain the theory of machine learning and neural networks, using vivid examples such as medieval man’s foot length, SIRI voice recognition, and blind dates.
2-1 Introduction to Machine learning
2-2 Introduction to neural networks
2-3 Neural network training
Chapter 3 introduction to tensorflow.js
Tensorflow.js is the core framework of this course. This chapter will help you understand your Arsenal before you do it. Why do you need Tensor
3-1 Tensorflow. Introduction of js
3-2 installation Tensoflow. Js
3-3 why do you need Tensor
Chapter 4 linear regression
This chapter will take you through the development and training of your first neural network model, which has only one neuron, but is the beginning of your machine learning journey!
4-1 Introduction to linear regression task
4-2 Preparation and visualization of training data and trial
4-3 Definition of model structure: a neural network composed of single neurons in a single layer
4-4 Loss function: mean square error
4-5 Optimizer: Stochastic gradient descent
4-6 Training model and visualization of training process
4-7 make predictions
Chapter 5 normalization
Nine to one… Wait, we’re not working on an abacus, we’re working on alchemy! This chapter will illustrate the alchemy of normalization using height and weight prediction as an example.
5-1 Introduction to normalized tasks
5-2 Normalized training data
5-3 Training, prediction, inverse normalization
Chapter 6 logistic regression
The task was to develop a neural network that could separate two kinds of points on a plane!
6-1 Introduction to logistic regression tasks
6-2 Loading binary data
6-3 Define model structure: single neuron with activation function
6-4 Loss function: Log Loss
6-5 Training model and visualization of training process
6-6 make predictions
6-7 (optional) Binary data set generation function source analysis
Chapter 7 multilayer neural network
There are not so many simple problems in life. In the face of complex problems, we can develop a multi-layer neural network model with activation function and swing the “knife” in our hands to cut them.
7-1 Introduction to multi-layer neural network tasks
7-2 Load the XOR dataset
7-3 Define model structure: multi-layer neural network
7-4 Training model and prediction
Chapter 8 multi-classification
This chapter will take iris classification as an example and learn how to use softmax and cross entropy algorithms to make neural network carry out multiple classification
8-1 Task Overview, main steps, and prerequisites
8-2 Loading IRIS Data Set (Training set and Verification Set)
8-3 Define model structure: multi-layer neural network with Softmax
8-4 Training model: Cross entropy loss function and accuracy measurement
8-5 Multi-classification prediction method
8-6 (Optional) Source code analysis of IRIS dataset generating functions
8-7 (Optional) Source code analysis of IRIS dataset generating functions
Chapter 9 underfitting and overfitting
It’s time to learn alchemy best Practices again! By the end of this task, you can take a glance at the training image and determine whether it is underfitting or overfitting.
9-1 Brief introduction to under-fitting and over-fitting tasks
9-2 Load dichotomous datasets with noise
9-3 Poor fitting demonstrated using simple neural networks
9-4 The fitting was demonstrated using complex neural networks
9-5 Overfitting response methods: early stop method, weight attenuation, discard method
Chapter 10 uses convolutional neural network (CNN) to recognize handwritten numbers
This chapter will first use a lot of animation to explain the theory of convolutional neural network, and then use JS to build and train it! Start building your first deep learning model!
10-1 Introduction to the task of handwritten numerals recognition using convolutional Neural Networks
10-2 Loading MNIST data set
10-3 Definition of model structure: convolutional neural network
10-4 Training model
10-5 for prediction
Chapter 11 uses pre-training model to classify images
Take the convolutional neural network model trained by others and use it directly! Bring people also need to learn oh!
11-1 Introduction to image classification tasks using pre-training models
11-2 Load the MobileNet model
11-3 for prediction
Chapter 12 Image classifier based on transfer learning: Trademark recognition
This chapter will use trademark recognition as an example to explain how to use transfer learning to classify images more efficiently. After learning this chapter, you can develop a variety of games and applications such as Paint me guess, flowers sorting, garbage sorting, emoji hunters and so on.
12-1 Image classifiers based on transfer learning: An introduction to trademark Recognition tasks
12-2 Load and visualize trademark training data
12-3 Defining model structure: truncation model + double-layer neural network
12-4 Model training under transfer learning
12-5 Model prediction under transfer learning
12-6 Model saving and loading
Chapter 13 uses the pre-training model for speech recognition
Voice recognition in the browser.
13-1 Introduction to speech recognition tasks using pre-training models
13-2 Loading the pre-trained speech recognition model
13-3 for speech recognition
Chapter 14 speech recognizer based on transfer learning: voice control wheel broadcast graph
This chapter will take you through the development of a remote voice control wheel map, and by the end of this chapter, you can build your own simple SIRI voice assistant!
14-1 Speech recognizer based on transfer learning: Voice control wheel broadcast graph
14-2 Collect Chinese voice training data in the browser
14-3 Training and prediction of speech recognition transfer learning
14-4 Saving and loading of voice training data
14-5 Voice control wheel broadcast diagram
Chapter 15. Python and JavaScript Models interweave
This chapter has covered the most useful and common techniques at work: converting a Python model to a JS model and deploying it to the browser. The JS model for fragmentation, compression, acceleration and other optimization transformation is also essential work!
15-1 Introduction to the Python and JavaScript Model interchange Task
15-2 Install Tensorflow.js Converter
15-3 Interaction between Python and JavaScript models
15-4 Interchange of JavaScript models: Sharding, quantization, acceleration
Chapter 16 Course Summary
Review the course as a whole.
16-1 – Review and summary
This course is over
六四运动