Teach on Udemy

Turn what you know into an opportunity and reach millions around the world.

Learn More

Your cart is empty.

Keep shopping

Regression in Angular using TensorFlow.js

Name: Regression in Angular using TensorFlow.js
Rating: 2.7 (7 reviews)

Learning to build nonlinear/linear regression models for datasets using machine learning in Typescript (applied math)

Created byJorge Guerra Pires, TensorFlow.js Academy

Last updated 5/2025

English

English [Auto],

What you'll learn

Creating linear/nonlinear regression models in TensorFlowjs
Basics of training a machine learning model
Building smart apps in Angular
Basics of linear/nonlinear regression
How to use Angular for scientific computation based on machine learning
Coding in TypeScript by example, basics, hands-on section

Course content

6 sections • 26 lectures • 3h 2m total length

Tips on learning and reusing code code snippet9:17
Attachment
Getting started with Angular, basics on installation;
GitBook
Organization of the course
PS. For each lecture, please, check the attached GitBook for more information
Interpreting the training graph14:56
Note. please, pay attention to the notes on the videos, just click on the note, and you will directed to the point on the vídeo.
Getting to know our AI-based Assistant13:50
Presenting the new Udemy AI-based assistant. You can interact and learn about the course. It is an awesome assistant, always available.
You do not neet to buy the course to see how the assistant works.
Getting to know the course

Introduction5:16
Attachment
Slides
Computational Thinking (ebook on Amazon)
How useful are simple regression models?5:00
Assess the value of simple linear regression models, focusing on bias and slope as learning parameters. The lecture contrasts analytical solutions with neural networks, recommending simple methods and packages.
Guiding your model: Absolute Error Summation vs. Mean Squared Error4:44
Explore how loss functions guide regression models in TensorFlow.js, comparing absolute error summation and mean squared error, and learn when to prefer each based on cost of mistakes.
How useful are simple regression models?2:05
Compare linear regression with simple models, highlighting speed and simplicity, and advise using classical mathematical models over TensorFlow.js when theory applies, while machine learning suits adaptable environments.
The mechanics of teaching a neural model: validation vs. training5:20

Initial words4:25
Compare linear regression's strong theory with artificial neural networks' flexibility for real-world regression, and learn when Gaussian noise assumptions and no parametric statistics favor non-linear models.
Back to the future: imagine you could send a machine learning model backward1:43
Attachment
article where I explain better the real-state model, the one I would send back into the past, from the future
Misconception: machine learning does not understand what it is doing4:35
Machine learning concept: features5:54
Explore how features influence house price predictions, assess feature correlations and independence, and decide between linear and nonlinear regression with PCA when needed, using TensorFlow.js.

Preparing the environment (with the help of chatGPT)5:11
Attachment
Getting you environment ready for Angular
Fixing initial issues, and we are good to go!7:52
Install TensorFlow.js in an Angular app to build our regression model and keep it running. Address common import errors by adjusting TypeScript typing and configuration, then restart the server.
Running a simple model, just to get familiar with TensorFlow.js10:16
Objetive ignorace: mechanical judgment vs. clinical judgment1:06
Examine objective ignorance and the contrast between mechanical and clinical judgment, noting models can’t know all data and must adapt to dynamic environments like the stock market.

Uploading our dataset from Google Spreadsheet11:13
Attachment.
I have attached the spreadsheet, in case of any problem using the link to the Google Sheet.
Taking a look on our dataset5:54
visualize helps you inspect a large dataset by importing a visualization library, installing npm packages, and plotting the data either in a specified div or by default on app startup.
Getting to know our code for plotting11:06
Learn to plot regression data with TensorFlow.js by sampling data points, visualizing training and testing, and mapping temperature to input X and target Y in a scatter plot.

Initial information4:25
Upload data from Google Sheets or use a dataset for TensorFlow.js in an Angular app. Train a model to learn inner dynamics from limited samples and validate with 100 samples.
Getting familiar with our code11:49
Explore how loss functions shape training in TensorFlow.js, comparing mean squared error versus absolute value based loss, and observe training versus validation curves to understand generalization and overfitting.
Learning to read the training curves11:49
Explore how loss functions guide training in TensorFlow.js, comparing mean squared error and mean absolute error, and read training and validation curves to assess convergence and generalization.
Batch size on the training curve2:12
Set the batch size to control how much data the model sees at once, speeding up learning and stabilizing training, especially with large datasets.
Add features to the model
Improving the sampling algorithm
Creating tensors from our dataset for training3:34
Transform features into tensors to prepare data for training with TensorFlow.js, constructing input feature tensors and target tensors, while coding generically for different features.
Final settings for finally training our model4:03
Build a regression model in TensorFlow.js with one input neuron and bias. Note no activation and explore unit counts, then compile and prepare the model for training.
Finally our hard work pays off: training the model9:36
Attachment
Full repository (GitHub)
Create a test graph

Requirements

I have tried to explain everything, but Angular knowledge may be advantageous, and a machine learning knowledge also may be useful.
All the tools used are open access, free of charge. You just need a computer with web browser

Description

Feature: this course uses an AI-assistant that you will support all the time. It is a talking assistant, you make your questions and it will reply back, a good conversation! You do not neet to buy the course to see how the assistant works.

Data Science is all about finding information/knowledge from datasets. One very powerful approach is using linear/nonlinear models, called regression; we are not going to see logistic regression on this course, see my other course “Machine learning in Angular”. Linear models, the example we are going to work on, even though they are limited, they still can delivery something if the datasets have a linear tendency. In TensorFlow.js, shifting to nonlinear models is as easy as changing a single parameter; that is all.

TensorFlow.js makes it possible to build simple models to complex ones (e.g., deep learning) using the same notation, it is magic! It is possible thanks to how TensorFlow.js was build, and the power of tensor, where TensorFlow.js is build on.

On this course, we use Angular as framework, coding environment, and TensorFlow.js as the library for creating a machine learning based regression model.

What is Angular??

Angular is a framework, designed by the Google Team, and it has been widely used to design sites. Essentially, it is a framework to create frontends, based on TypeScript. In layman's terms: the page you see and interact on your web browser. It is perfect to be used alongside TensorFlow.js for offering also the advantage of not sending information to the server: you can build everything with Angular, they even have their own server call Angular Universal. As they like to say, Angular comes with batteries.

What is TensoFlow.js??

TensorFlow.js is a JavaScript-based library for deep learning, based on the classical TensorFlow, written in Python; you can also do simple learning machine, some simple mathematical operations with tensors and so on. There are several reasons for using TensorFlow.js instead of Python, and I hope to come back to that in the future.

A nice point, it is possible to transform models in both directions: TensorFlow.js <-> TensorFlow. Even their notations are alike.

We are going to build a linear regression model using TensorFlow.js in Angular. We are also going to learn about machine learning, and Angular!

External resource

- GitBook

- GitHub

- Articles

Update 18/08/23. in response to student feedback, I am upgrading the course, adding more resources.

Who this course is for:

JavaScript programmers wanting to learn machine learning
Machine learning practitioners wanting to learn web coding
Angular coders wanting to add intelligence to their apps
Programmers to start on Angular using practical problems

Regression in Angular using TensorFlow.js

What you'll learn

Explore related topics

Course content

Tips on coding4 lectures • 38min

Discussing a simple regression model: when we may have to decide the best model5 lectures • 22min

Nonlinear regression and neural networks4 lectures • 17min

Building a linear regression to predict perceived temperature4 lectures • 24min

Uploading and visualizing our dataset3 lectures • 28min

Creating and training our model in TensorFlow.js7 lectures • 47min

Requirements

Description

Who this course is for: