T-1: Varieties of human-like AI

Saturday, 27 August, 10:30 - 12:15 Pacific Time (UTC -7)
Location: Grand Ballroom A
Tutorial

This tutorial aims to enable participants to learn the practical basics of designing and running reinforcement learning experiments for cognitive and behavioral neuroscience. To do so, the tutorial will utilize the recently released open source Neuro-Nav python toolkit. Instead of needing to install the toolkit locally, participants will be able to directly follow along by running the examples using provided google colab notebooks from their web browsers without the need to install any additional software.

The tutorial will first cover the basics of the toolkit and of framing reinforcement learning problems. It will then walk through a few examples of reproducing existing cognitive neuroscience results in the literature using Neuro-Nav. Next it will cover using Neuro-Nav to design and run a simple novel study comparing the behavior of different reinforcement learning algorithms. The final section of the tutorial will involve extending research into the multi-agent domain, and the study of behavior which exists within larger social structures.

Ida Momennejad

Ida Momennejad

Microsoft Research

Arthur Juliani

Arthur Juliani

Microsoft Research