Cloud IoT Real Time Learning Eng: Computers Teach Themselves (Bill Schreiber)
Details
Cloud IoT Real Time Reinforcement Learning Engine: Computers that Teach Themselves
Description
Developing Machine Learning and Artificial Intelligence applications for the Internet Of Things presents design challenges since it needs to work with devices that have limited resources and connectivity. Machine Learning and Artificial Intelligence implementations rely on lots of memory and processing power so an IoT AI implementation lends itself to a distributed approach. The question is where to draw that line to decouple Machine Learning or Artificial Intelligence algorithms from the device control and maximize reuse.
My talk will delve into this architecture topic as well as show two demos using my library. I will show an example where in real time the computer teaches itself Pong, and another where a robot teaches itself the best strategy to do if it runs into anything. Understanding what reinforcement learning can and cannot do and how to think about what problems it can solve is important to the businesses we help.
My hobby is robotics and one of my interests is computers that teach themselves. I started playing with this idea originally in 2014 and have worked on this project on and off (mostly off) refining its implementation in my spare time. There have been many iterations, starts and stops, two complete rewrites from scratch, and an embarrassing amount of development time to get it right, but the end result is a scalable reinforcement learning engine that allows a number of robots or computers to learn from experience in real time. It uses an XML interface to configure what it teaches so no programming is required for a device or computer to use it.
Bio
Bill Schreiber
I am a software developer with 18 years of programming experience. I am presently a principal developer at RSA in Bedford MA where my group develops, maintains and augments the systems that run the company in their IT department. I am passionate about development, security and always looking to learn, grow and become a better person and engineer.
Venue and Food
We meet at Magenic (see address above) at 6-8 pm. As usual, there will be pizza and sodas provided. Please RSVP through this site if you will be attending.