Artem Kirsanov | Can We Build an Artificial Hippocampus? @ArtemKirsanov | Uploaded 1 year ago | Updated 2 hours ago
To try everything Brilliant has to offer—free—for a full 30 days, visit http://brilliant.org/ArtemKirsanov
The first 200 of you will get 20% off Brilliant’s annual premium subscription.
My name is Artem, I'm a computational neuroscience student and researcher. In this video we discuss the Tolman-Eichenbaum Machine – a computational model of a hippocampal formation, which unifies memory and spatial navigation under a common framework.
Patreon: patreon.com/artemkirsanov
Twitter: twitter.com/ArtemKRSV
OUTLINE:
00:00 - Introduction
01:13 - Motivation: Agents, Rewards and Actions
03:17 - Prediction Problem
05:58 - Model architecture
06:46 - Position module
07:40 - Memory module
08:57 - Running TEM step-by-step
11:37 - Model performance
13:33 - Cellular representations
17:48 - TEM predicts remapping laws
19:37 - Recap and Acknowledgments
20:53 - TEM as a Transformer network
21:55 - Brilliant
23:19 - Outro
REFERENCES:
1. Whittington, J. C. R. et al. The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation. Cell 183, 1249-1263.e23 (2020).
2. Whittington, J. C. R., Warren, J. & Behrens, T. E. J. Relating transformers to models and neural representations of the hippocampal formation. Preprint at http://arxiv.org/abs/2112.04035 (2022).
3. Whittington, J. C. R., McCaffary, D., Bakermans, J. J. W. & Behrens, T. E. J. How to build a cognitive map. Nat Neurosci 25, 1257–1272 (2022).
CREDITS:
Special thanks to Crimson Ghoul for providing English subtitles!
Icons by biorender.com and freepik.com
Brain 3D models were created with Blender software using publicly available BrainGlobe atlases (brainglobe.info/atlas-api)
Animations were made using open-source Python packages Matplotlib and RatInABox ( github.com/TomGeorge1234/RatInABox )
Rat free 3D model: skfb.ly/oEq7y
This video was sponsored by Brilliant
To try everything Brilliant has to offer—free—for a full 30 days, visit http://brilliant.org/ArtemKirsanov
The first 200 of you will get 20% off Brilliant’s annual premium subscription.
My name is Artem, I'm a computational neuroscience student and researcher. In this video we discuss the Tolman-Eichenbaum Machine – a computational model of a hippocampal formation, which unifies memory and spatial navigation under a common framework.
Patreon: patreon.com/artemkirsanov
Twitter: twitter.com/ArtemKRSV
OUTLINE:
00:00 - Introduction
01:13 - Motivation: Agents, Rewards and Actions
03:17 - Prediction Problem
05:58 - Model architecture
06:46 - Position module
07:40 - Memory module
08:57 - Running TEM step-by-step
11:37 - Model performance
13:33 - Cellular representations
17:48 - TEM predicts remapping laws
19:37 - Recap and Acknowledgments
20:53 - TEM as a Transformer network
21:55 - Brilliant
23:19 - Outro
REFERENCES:
1. Whittington, J. C. R. et al. The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation. Cell 183, 1249-1263.e23 (2020).
2. Whittington, J. C. R., Warren, J. & Behrens, T. E. J. Relating transformers to models and neural representations of the hippocampal formation. Preprint at http://arxiv.org/abs/2112.04035 (2022).
3. Whittington, J. C. R., McCaffary, D., Bakermans, J. J. W. & Behrens, T. E. J. How to build a cognitive map. Nat Neurosci 25, 1257–1272 (2022).
CREDITS:
Special thanks to Crimson Ghoul for providing English subtitles!
Icons by biorender.com and freepik.com
Brain 3D models were created with Blender software using publicly available BrainGlobe atlases (brainglobe.info/atlas-api)
Animations were made using open-source Python packages Matplotlib and RatInABox ( github.com/TomGeorge1234/RatInABox )
Rat free 3D model: skfb.ly/oEq7y
This video was sponsored by Brilliant
