Distributed AI in the BrowserLaajuus (5 ECTS)

Objective

Knowledge and understanding:

The students will know some basic concept and some algorithms of machine learning for regression and classification and they will understand how to implement them in Keras.
They also will learn how to distribute these algorithms along multiple nodes and will learn how to elegantly and ergonomically convey these algorithms, from design to execution, in a browser using JavaScript and other programming languages.

Skills:
The students will be able to rapidly prototype distributed, browser-based machine learning ideas.

Content

Motivation:
As innovative projects like Deep Brain Chain, SONM, XNOR and NNVM emerge, there is an increasingly rich ecology of out-of-cloud approaches to artificial intelligence in both computing and inference. This course sits at the nexus of these two trends, combining distributed computing and edge computing in a browser environment. As a powerful tool for sharing and displaying media, the browser is becoming the de facto playground for prototyping novel distributed computing and edge computing ideas.

Core content level:
Introduction to distributed algorithms and edge computing
Introduction to Keras for implementation of machine learning algorithms
Exporting a Keras model to run in a browser
Representation of a machine learning model’s training in the browser
Representation of a machine learning model’s construction in the browser
Distributed machine learning model computation across multiple browsers
Controlling the training and inference of a distributed machine learning model in real time 1
- Design
Controlling the training and inference of a distributed machine learning model in real time 2
- Implementation and Execution

Additional content:
Showcasing some student distributed browser-based machine learning projects

Qualifications

Basic mathematics, for example, linear algebra, vector and matrix operations, linear combination, basic multivariate calculus. Intermediate web experience, such as web design, coding websites and familiarity with popular frameworks like React or Angular. It is preferable to have basic programming skills in python and to be able to use basic data structures and algorithms.

Assessment criteria, satisfactory (1)

Student understands fundamentals of machine learning and representation of machine learning in web, and is able to utilize existing tools.

Assessment criteria, good (3)

In addition to satisfactory level, the student is able to choose an appropriate machine learning model for a given problem, and implement an appropriate UX and UI to interact with the model via web browser.

Assessment criteria, excellent (5)

In addition to good level, the student is able to design a unique end-to-end web based machine learning solution for a given problem.

Assessment criteria, approved/failed

Student understands fundamentals of machine learning and representation of machine learning in web, and is able to utilize existing tools.