JSFoo: round the year submissions

JSFoo: round the year submissions

Submit talks on JavaScript and full stack engineering round the year

Submit talks and workshops on:

  1. Full stack engineering
  2. JavaScript frameworks
  3. Architecture approaches and case studies
  4. JavaScript and IoT
  5. Web development
  6. JavaScript for hobbyists

We will find a forum to place your talk, any time during the year.

For more information, contact jsfoo.editorial@hasgeek.com or call 7676332020

Hosted by

JSFoo is a forum for discussing UI engineering; fullstack development; web applications engineering, performance, security and design; accessibility; and latest developments in #JavaScript. Follow JSFoo on Twitter more

Girimurugan Natarajan

@girimurugan

Machine Learning on Browser using TensorFlow.js

Submitted Apr 20, 2018

When we talk about machine learning it is usually languages like Python or statisticaly packages like R which comes to our mind. JavaScript could be the last to come up in the list, since it is usually associated with web development or server side Node.js.

This is about to change, not all machine learning problem are big require heavy compute powers, some of them could be solved by processing power of the browsers. TensorFlow.js comes to help us here.

Outline

Why JavaScript for Machine Learning

  • Not all machine learning problems require high-end computation and solutions
  • JavaScript community can adopt machine learning faster
  • Some of the machine learning problems can be solved in UI layer

Covers both low-level & high-level machine learning

  • What is the difference between this classification
  • When to use which approach

API to consume pre-trained models

  • Explanation on the need for APIs and how to consume Kera models in TensorFlow.js

Examples of Machine learning (data)

  • Discuss a simple example use case of linear progression using JavaScript linear algebra library

Example of Machine learning on image recognition

  • Discuss an example of image recognition to extract data from images with pre-trained model

Adoption & Future in store

  • Performance of the TensorFlow.js in browser is 1.5X (smaller model) to 15X (huge data sets) againts TensorFlow in Python. So, use is judiacially
  • TensorFlow.js is not supported in Node.js yet, but we can expect the support in near future due to popularity of Node.js
  • Great fit for simple machine learning use case which can be solved in UI client side

Requirements

  • no specific requirement -

Speaker bio

Girimurugan has 15 years of experience in enterprise software development. Has worked with SAP for more than a decade on various web applications and currently contributes for WalmartLabs in Bangalore. Girimurugan has extensive experience on developing & designing web application of enterprise scale. He is currently focusing on solving enterprise problems with Conversational User Experience & Machine Learning.

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Hosted by

JSFoo is a forum for discussing UI engineering; fullstack development; web applications engineering, performance, security and design; accessibility; and latest developments in #JavaScript. Follow JSFoo on Twitter more