Machine Learning and Deep Learning with JavaScript

Name: Machine Learning and Deep Learning with JavaScript
Rating: 2.9 (8 reviews)

Learn Machine Learning and Deep Learning from scratch using JavaScript and Tensorflow.js with hands-on projects

Created byPackt Publishing

Last updated 6/2019

English

What you'll learn

Get acquainted with machine learning and deep learning capabilities using JavaScript and understand the JavaScript Machine Learning ecosystem
Learn JavaScript libraries to build neural network models
Know how to decide, analyze, and make predictions from real-world data
Solve real-world problems such as predicting mental health issues
Use clustering algorithms to understand customer behavior and categorize customers
Train your machine learning models to work with different kinds of data
Work with powerful algorithms using the pre-written libraries in Python
Build deep learning models with TensorFlow .js and practice on realistic datasets

Course content

3 sections • 71 lectures • 6h 27m total length

The Course Overview1:19
This video provides an overview of the entire course.
Introduction to Machine Learning6:35
This video aims to teach the viewer what Machine Learning is, so that the concepts of the rest of the course can be understood.
• Define what Machine Learning Is
• Discuss the different types of machine learning
• Give an example of each type of Machine Learning Technique
Tour of the JavaScript Machine Learning Landscape4:04
In this video, we will get comfortable with the idea of using JavaScript for Machine Learning.
• Analyze the advantages of using JavaScript for Machine Learning
• Look at Machine Learning Tools library
• Understand the TensorFlow.js library and STDlib library
Setting Up Our Machine Learning Environment7:56
In this video, we will look at how to set up our environment for learning/using Machine Learning in Javascript for the rest of the course.
• Set up our Server
• Install the visual code studio
• Use the web browser to see our output
Understand Regression with Linear Regression6:36
We will understand what a Regression problem is and how Linear Regression can help us.
• Learn about regression as a form of supervised learning
• Look at an example of a Regression Problem
• See how Linear Regression can help us solve the problem
Understanding How Linear Regression Works5:24
This video aims to teach the viewer about the core concepts in Regression as well as machine learning.
• Define what is a linear model
• Look at some of the fundamental concepts in machine learning, loss function and optimizer
• Understand the impact and importance of different learning rates
Predicting Salaries after College Using Linear Regression4:09
In this video, we will implement the linear regression code in JavaScript using the ML-Regression library.
• Go through the steps in linear regression
• Create the setup for executing JavaScript code
• Look at a JavaScript code to understand how linear regression works and how to implement it
Understand Classification with Logistic Regression4:10
This video goes into the details of classification.
• See what is a classification problem
• Learn about logistic regression
• Understand how the cost function is different from linear regression
Classifying Clothes Using Logistic Regression6:09
In this video, we will implement the logistic regression code in JavaScript.
• We will go through the steps in logistic regression
• Write the code to calculate loss function
• Look at the code to understand how logistic regression works and how to implement it
Model Evaluation3:30
We will learn about ways to design our data, so that we can improve performance.
• Learn about what we can and cannot change to improve performance
• Look at validation set as a proxy test set
• See how K-fold cross validation can be used for parameter tuning
Better Measures than Accuracy8:00
This video aims to teach the viewer about different ways we can measure the performance of a model.
• We see what are classification metrics
• We use the classification metrics to get a clearer picture about the model
• We learn about some of the regression metrics available to us
Understanding the Results2:20
In this video, we will understand how to interpret the performance measures.
• Understand bias and underfitting
• Understand variance and overfitting
• Learn about bias variance trade-off
Improving the Models2:55
This video goes into the details of techniques to further improve our model.
• Learn about regularization
• Know how to use regularization
• Look at another technique that improves our results called as feature scaling
What are Support Vector Machines?2:21
We will understand what Support Vector Machines are and how it is used.
• Learn about Support Vector Machines as a form of supervised learning
• Look at how support vector machines work
• Know how support vector machines work with the help of a graph
Using SVM Kernels to Transform Problems1:42
This video aims to teach the viewer about Kernels in Support Vector Machines.
• Understand what are non-linear problems
• Look at Kernels in SVM
• Use Kernels to transform problems so that we can use SVMs
Image Classifier Using SVM8:34
In this video, we will implement the SVM code in JavaScript using the LIBSVM library.
• Understand the problem
• Know a way to represent the features in a way SVM can understand
• Look at the JavaScript code
Making Better Decision with Decision Trees7:59
We will learn about decision trees in details.
• Learn about Decision Trees and see how they are used
• Look at the advantages of Decision Trees
• Understand the underlying methods behind Decision Trees
Combining Decision Trees to Make Better Predictions2:35
In this video, we will about ways to combine decision trees to get better models.
• Learn about ensemble learning
• Learn about Random Forest
• Get an understanding of how useful Random Forest are
Predicting Customer Churn Using Random Forests5:32
This video is all about using Random Forest to solve the complex problem of Customer Churn.
• Understand the problem of customer churn
• See how Random Forest can help solve the problem
• Compare the results with Decision Trees
Introduction and Advantage of Unsupervised Learning2:39
We will get an introduction to unsupervised learning and look at some of its advantages.
• Understand what is unsupervised learning
• Look at some examples of unsupervised learning
Grouping Unlabeled Data in Meaningful Ways Using K-means Clustering5:03
This video aims to teach the viewer about clustering and k means clustering algorithm.
• Understand what is clustering
• Learn how to use k means clustering
• Know how to determine the optimal number of clusters
Using Principal Component Analysis to Speed-up Machine Learning Algorithms4:24
In this video, we will look at another unsupervised technique called as Principal Component Analysis.
• Learn about Principle Components
• Get an understanding of Eigenvectors and Eigenvalues
• Understand how PCA can be used for dimensionality reduction
Analyzing Plant Species Using K-means Clustering3:11
In this video, we will use K-means clustering in JavaScript.
• Look at the problem we are trying to solve
• Use K-means to solve the problem
• Compare different initialization techniques to compare the results
Introduction to Neural Networks1:46
We will get an introduction to Neural Networks.
• Understand what are Neural Networks
• See the structure of a simple neural network
• Look at Neural Network design of a Feedforward Neural Network
How a Neural Network Works6:18
This video aims to teach the viewer about the different parts of a Neural Network and its functions.
• Calculate the Neural Network Output
• See how to get the cost function
• Learn about Backpropagation as a way to update my weights
Neural Networks in Tensorflow.js4:15
In this video, we will understand how to use TensorFlow.js for Neural Networks.
• Get introduced to TensorFLow.js
• Learn about its fundamentals
• /use TensorFLow.js to build a Neural Network with just one neuron
Multiclass Classification Using TensorFlow.js4:49
In this video, we will build a Neural Network using TensorFlow to solve digit classification.
• Understand the problem as hand
• Look at a simple Neural Network architecture
• Use TensorFlow.js to solve the problem
Test your knowledge

The Course Overview1:39
This video provides an overview of the entire course.
Introduction to Machine Learning3:48
This video aims to teach the viewer what Machine Learning is, so that the concepts of the rest of the course can be understood.
Define what Machine Learning Is
Discuss the different types of machine learning
Give an example of each type of Machine Learning Technique
Getting Started with TensorFlow.js Using a Simple Example to Predict Weight3:38
In this video, we will get an introduction to TensorFlow and its concepts.
Learn about TensorFlow
Understand the fundamentals of TensorFlow
Look at a sample TensorFlow Code to understand its Flow
Setting Up Our Machine Learning Environment7:55
In this video, we will look at how to set up our environment for learning/using TensorFlow for the rest of the course.
Set up our Server
Install the visual code studio to write our JavaScript code
Use the web browser to see our TensorFlow output
Types of Supervised Learning3:20
In this video, we cover supervised learning in more details.
Understand what is supervised learning
Learn about regression as a form of supervised learning
Get to know classification as a form of supervised learning
Applying Regression6:30
This video aims to teach the viewer about the core concepts in Regression as well as machine learning.
Define what is a linear model
Look at some of the fundamental concepts in machine learning of loss function and optimizer
Understand the impact and importance of different learning rates
Predicting Salaries after College Using TensorFlow7:06
In this video, we will implement the linear regression code in TensorFlow.
Go through the steps in linear regression
Revisit the fundamentals of TensorFlow
Look at a TensorFlow Code to understand how linear regression works and how to implement it
Applying Classification3:49
This video goes into the details of classification.
Understand what is a classification problem
Learn about logistic regression
Analyze how the cost function is different from linear regression
Predicting Mental Health Issues Using Logistic Regression6:02
In this video, we will implement the logistic regression code in TensorFlow.
Go through the steps in logistic regression
Write the code to calculate loss function
Look at a TensorFlow Code to understand how logistic regression works and how to implement it
Understanding Simple Neural Networks2:03
In this video, we cover supervised learning in more details.
Start by discussing what is a neural network
Learn about deep neural networks
See what are the advantages of using deep neural networks
Concepts in Neural Network3:50
What are the core concepts in Neural Networks?
Learn about neurons
Get to know about a special type of neuron called as a sigmoid neuron
Understand what are activation functions and how the property of different neural networks change based on the activation function
Working with Deep Neural Networks5:28
What are the different steps taken during a deep neural network training?
Learn about gradient descent
Understand the ways to pass gradient up and down the neural network
Know about the bias term which helps in getting better results
Image Classification Using Neural Networks8:19
What are the details applied in neural networks?
Understand the problem
Create a simple neural network
Train the model and get results
Model Evaluation4:13
What are the techniques to measure model performance?
Start by defining the problem
Look at the trade-off between bias and variance
Learn about validation set to be used while testing our model
Better Measures than Accuracy5:13
What are some of the better ways of getting model performance?
Learn about classification metrics
Get to know about confusion metrics
Understand what are regression metrics
Improving the Models6:14
What are the different ways to improve our model?
Analyse the cross-entropy cost function
Get an understanding of Regularization
See some other regularization techniques like Dropout and ensemble learning
Optimizing the Models4:12
What is optimization? And what are the problems involved like vanishing gradient problem?
Learn about batching the data into different sized groups
Know about the vanishing/ exploding gradient problem
Get to know one of the ways to prevent vanishing gradient problem
Using High-Level Layers API to Construct Neural Networks7:35
What are the various layers of API?
Revisit neural networks to get a comparison with layers
Get to know about the core API which we have been using so far
We learn about the layers API and see some of its advantages
Building Advanced Neural Networks with Layers Easily4:40
How to build advanced neural networks? What are Convolutional Neural Networks?
Start by understanding weight sharing
Learn about convolution neural network which uses the idea of weight sharing
Understand what is max pooling and average pooling
Detecting Digits Using Layers8:13
We go into a hands-on example of using layers.
Define the problem of hand written digit classification
Learn how to use layers
Build a convolutional neural network with just layers
Building A Classifier Using Layers4:22
We look at another hands on example of using layers.
See what is the classification problem
Build our classifier using layers
Use layers to train and test our model
Importing a Keras Model into TensorFlow.js3:49
How to import Keras model into TensorFlow?
Learn how TensorFlow.js python convert works
Get to know about the Python API to export directly to TF.js
Understand ways to load the model into TensorFlow.js
Saving and Loading TensorFlow Models6:17
How to save and load our pre-trained model at different locations?
Learn about the importance of saving our model
Know different ways in which we can save our model
See different ways in which we can load the models
Importing TensorFlow SavedModel into TensorFlow.js4:22
In this video, we will learn about ways to load TensorFlow saved models into TensorFlow.js
See how to convert TensorFlow models into TensorFlow.js
Know how to load the TensorFlow model
Learn about running the TensorFlow model
Playing PAC-MAN Using a Webcam7:16
In this video, we will learn how to use our knowledge so far and create a game which uses image classification
Know about how the data is collected for the game
See how the model is trained on new data
Understand how to use transfer learning to save training time and effort
Test your knowledge

The Course Overview6:48
This video provides an overview of the entire course.
What Makes Deep Learning in JavaScript Special?2:38
The goal of this video is to highlight some main benefits of using deep learning methods using JavaScript.
Why a web browser can be a great tool to learn deep learning
How you can save a lot of time by learning deep learning
See new use cases for deep learning using JS
Getting Started with TensorFlow.js5:56
Learn how to quickly get started with TensorFlow.js.
Create an HTML page
Insert the script tag at the top of the head section with src pointing at the TensorFlow.js library
Make sure you can access the tf variable, which contains access to the TensorFlow.js API
Loading Pre-Trained CNN and LSTM Models7:23
Where and how to get a pre-trained model for sentiment analysis?
Define the right URLs for both the model and metadata
Test both URLs
Load the model, then load the metadata, and set important values for later use
Preparing a New Text for Sentiment Analysis7:57
How to prepare a new text for use with our pre-trained models?
Remove all the unnecessary characters and turn the text into an array of words
Use a special array (buffer) to encode words according to the metadata
Turn buffer into another special array called tensor
Using Loaded Model for Real-Time Text Analysis10:08
How to do sentiment analysis on an actual text in practice?
Define clearly where the input text is and where the output sentiment score should be shown
Define the necessary callback functions to work with the loaded model
Get and prepare the input text, do the analysis with the model, get the sentiment score, format it, and show it where you need it
Loading a Set of Pre-Trained CNN Models for Emotion Detection in Photos9:13
Learn how to load a set of pre-trained models for emotion detection.
Place the right models in the right locations
Run an HTTP server to place models from the existing location
Configure both FaceDetector and EmotionDetector to use models from the right location
Preparing a New Image for Analysis6:51
Discover how to prepare a new image for both face detection and emotion detection.
Learn how load an image to canvas for easy processing
Us a Face API JS to detect faces in an image
Learn how to detect a variety of emotions and their probabilities from detected faces
Using Our Models for Photo Emotion Detection3:57
Learn how to add information about detected emotions, both on an image and separately in a table.
Find which of the detected emotions is most probable
Mark a detected face and label it with the most probable emotion
Process each supported emotion and create a table with probabilities for each
Loading a Pre-Trained CNN Model for Voice Emotion Detection3:34
Learn where to place the model files and what are they, learn how use tf.load to load a specific model file
Download and put the model files into the right location
Load model files asynchronously
Check for errors
Preparing a New Audio Sample for Analysis11:15
Learn how to encode your audio sample for audio detection
Extract MFCC features from an audio sample
Put those features first into a buffer and then a tensor
Reshape the tensor so it can match the model
Using the Loaded CNN Model for Detecting Emotions in Speech4:02
How to do the emotion detection/prediction using our model
Run model.predict on prepared tensor
Convert output to a set of indexes related to emotions
Pick up the most probable index
Create a New Model Based on a Pre-Trained CNN Model5:52
How to extract key information from pre trained model and used to create a new model?
Load pre trained model
Find the last Convolutional layer, extract it and use it to create the new “seed” model with the same inputs as the pre trained model
Create the second model, configure it so that it can access the inputs from the first one and do the classification
Getting and Preparing a New Audio Sample for Training and Testing5:27
How to use the first “seed” model to prepare the data
Extract MFCC features from audio samples
Convert the data to the format understood by the network
Use this data to make predictions using the “seed” model and in that way complete the data preparation process
Training and Testing the New Model9:28
How to easily explore how to train, validate and test my transfer learning model?
Run a http server in the source directory for this section to run a training app
Understand the key aspects of training process using training app
Run the training app, interpret results
Getting and Preparing Audio Sample9:44
How to convert audio samples for training our CNN?
Find the free data that can be used in our project
Identify the type of features that we will be using to represent emotion in audio samples
Learn how to load and prepare converted data to use with our CNN
Building a CNN Model for Emotion Detection8:41
How to build a CNN that could perform emotion detection
Identify which TensorFlow API to use for building a model
Understand the inputs and outputs of the network, add the rest of necessary network’s layers
Choose the right optimizer, compile the network with the right loss function and metrics
Training and Testing the Model9:10
Learn how to train, validate and test the model
Understanding the idea behind splitting data into three parts - training, validation and testing
Learn how to use TensorFlow JS API for training, validation and testing
Run the small web app to see how to train our model, understand training charts and variables
Using Trained CNN Model on New Audio Samples4:59
Learn how to download trained model and use it (Similar to Section 4)
Prepare the new audio samples for detection
Predict the emotion using our trained model
Interpret and show the results of model prediction
Test your knowledge

Requirements

Working knowledge of JavaScript is required.

Description

Machine learning and Deep Learning have been gaining immense traction lately, but until now JavaScript developers have not been able to take advantage of it due to the steep learning curve involved in learning a new language. Here comes a browser based JavaScript library, TensorFlow.js to your rescue using which you can train and deploy machine learning models entirely in the browser. If you’re a JavaScript developer who wants to enter the field ML and DL using TensorFlow.js, then this course is for you.

This course takes a step by step approach to teach you how to use JavaScript library, TensorFlow.js for performing machine learning and deep learning on a day-to-day basis. Beginning with an introduction to machine learning, you will learn how to create machine learning models, neural networks, and deep learning models with practical projects. You will then learn how to include a pre-trained model into your own web application to detect human emotions based on pictures and voices. You will also learn how to modify a pre-trained model to train the emotional detector from scratch using your own data.

Towards the end of this course, you will be able to implement Machine Learning and Deep Learning for your own projects using JavaScript and the TensorFlow.js library.

Meet Your Expert(s):

We have the best work of the following esteemed author(s) to ensure that your learning journey is smooth:

Arish Ali started his machine learning journey 5 years ago by winning an all-India machine learning competition conducted by IISC and Microsoft. He was a data scientist at Mu Sigma, one of the biggest analytics firms in India. He has worked on some cutting-edge problems involved in multi-touch attribution modeling, market mix modeling, and Deep Neural Networks. He has also been an Adjunct faculty for Predictive Business Analytics at the Bridge School of Management, which along with Northwestern University (SPS) offers a course in Predictive Business Analytics. He has also worked at a mental health startup called Bemo as an AI developer where his role was to help automate the therapy provided to users and make it more personalized. He is currently the CEO at Neurofy Pvt Ltd, a people analytics startup.
Jakub Konczyk has done programming professionally since 1995. He is a Python and Django expert and has been involved in building complex systems since 2006. He loves to simplify and teach programming subjects and share them with others. He first discovered Machine Learning when he was trying to predict real estate prices in one of the early stages startups he was involved in. He failed miserably. Then he discovered a much more practical way to learn Machine Learning that he would like to share with you in this course. It boils down to Keep it simple!

Who this course is for:

This course is for JavaScript developers interested in Machine Learning and Deep learning. This course is also for data analysts and data scientists who want to explore the possibilities of Machine Learning and Deep Learning using JavaScript.

Machine Learning and Deep Learning with JavaScript

What you'll learn

Explore related topics

Course content

Hands-On Machine Learning using JavaScript27 lectures • 2hr 4min

Hands-On Machine Learning with TensorFlow.js25 lectures • 2hr 10min

Deep Learning Projects with JavaScript19 lectures • 2hr 13min

Requirements

Description

Who this course is for: