Timezone: »

Machine Learning in the Browser
Ted Meeds · Remco Hendriks · Said Al Faraby · Magiel Bruntink · Max Welling

Tue Dec 04:00 PM -- 08:59 PM PST @ Level 2, room 230B
Event URL: http://software-engineering-amsterdam.github.io/MLitB/ »

The ubiquity of the browser as a computational engine makes it an ideal platform for the development of massively distributed and collaborative machine learning. This will be demonstrated with our software framework written in Javascript, called MLitB. Our first use case is capable of performing large-scale distributed stochastic gradient descent learning of a DNN using heterogeneous compute devices, including, but not limited to, smart phones and tablets. Demo participants will be able to collaboratively train a DNN for image classification during the poster session and then use the model on their phones.

Implications of this software paradigm include cheap, distributed computing on a massive scale, truly collaborative and reproducible machine learning research, publicly accessible machine learning models and algorithms, real privacy preserving computation, field research, green computing, and more. This software is open source and freely available.

Author Information

Ted Meeds (University of Amsterdam)
Remco Hendriks (University of Amsterdam)
Said Al Faraby (University of Amsterdam)
Magiel Bruntink (University of Amsterdam)
Max Welling (University of Amsterdam / Qualcomm AI Research)

More from the Same Authors