Explore the Tolman-Eichenbaum Machine, a computational model unifying memory and spatial navigation in the hippocampal formation. Delve into the model's architecture, including position and memory modules, and understand its step-by-step operation. Examine the model's performance, cellular representations, and its ability to predict remapping laws. Learn how this framework relates to Transformer networks and gain insights into cognitive map building. Discover the connections between computational neuroscience, artificial intelligence, and our understanding of memory and spatial navigation in this informative 24-minute video lecture.
A Unifying Framework for Memory and Abstraction
Overview
Syllabus
- Introduction
- Motivation: Agents, Rewards and Actions
- Prediction Problem
- Model architecture
- Position module
- Memory module
- Running TEM step-by-step
- Model performance
- Cellular representations
- TEM predicts remapping laws
- Recap and Acknowledgments
- TEM as a Transformer network
- Brilliant
- Outro
Taught by
Artem Kirsanov
