Multi-Task Reinforcement Learning with Dendrites / Embodied AI - May 19, 2021
Numenta’s research intern Akash Velu explores how dendritic networks can be used in multi-task reinforcement learning through the MetaWorld Environment. Akash first gives an overview of multi-task RL and highlights a few challenges. He then presents and evaluates approaches he’s conducted on how to achieve strong performance with dendritic networks in multi-task RL. He goes over his progress and the challenges he’s faced in this project.
In the second half, visiting scientist Viviane Clay gives a progress on her project about learning meaningful representations through self-supervised exploration and interaction. In her experiment, she addresses how a model can learn without rewards by looking at a curiosity-driven exploration and exploration by disagreement framework outlined in two papers by Pathak et al.
0:00 Multi-Task Reinforcement Learning with Dendrites
42:15 Embodied AI
Papers from Viviane’s presentation:
➤ “Curiosity-driven Exploration by Self-supervised Prediction” (2017) by Pathak et al: https://arxiv.org/abs/1705.05363
➤ “Self-Supervised Exploration via Disagreement” (2019) by Pathak et al: https://arxiv.org/abs/1906.04161
➤ "Learning Sparse and Meaningful Representations Through Embodiment“ (2020) by Clay et al: https://doi.org/10.1016/j.neunet.2020...
➤ "Fast Concept Mapping: The Emergence of Human Abilities in Artificial Neural Networks when Learning Embodied and Self-Supervised“ (2021) by Clay et al: https://arxiv.org/abs/2102.02153
- - - -
Numenta is leading the new era of machine intelligence. Our deep experience in theoretical neuroscience research has led to tremendous discoveries on how the brain works. We have developed a framework called the Thousand Brains Theory of Intelligence that will be fundamental to advancing the state of artificial intelligence and machine learning. By applying this theory to existing deep learning systems, we are addressing today’s bottlenecks while enabling tomorrow’s applications.
Subscribe to our News Digest for the latest news about neuroscience and artificial intelligence:
https://tinyurl.com/NumentaNewsDigest
Subscribe to our Newsletter for the latest Numenta updates:
https://tinyurl.com/NumentaNewsletter
Our Social Media:
/ numenta
/ officialnumenta
/ numenta
Our Open Source Resources:
https://github.com/numenta
https://discourse.numenta.org/
Our Website:
https://numenta.com/
