CS-479 Learning in neural networks

Brain-style learning in Neural Networks: Learning algorithms of the brain. 

The course focuses on Reinforcement Learning and Self-supervised/unsupervised learning paradigms with an emphasis on the foundations of learning algorithms in biology.

EPFL, Lausanne, Switzerland

1. Why BackProp is biologically not plausible. Two-factor and three-factor rules. Bandit learning

REINFORCEMENT LEARNING (RL)

2. Three-factor rules for reward-based learning (Reinforcement Learning 1)

3. Three-factor rules for TD learning: SARSA and eligibility traces (Reinforcement Learning 2)

4. Neuromorphic hardware and in-memory computing (Application 1/external speaker)

5. Policy gradient (Reinforcement Learning 3)

6. Actor-critic networks (Reinforcement Learning 4)

7. Reinforcement learning in the brain (Reinforcement Learning 5)

SELF-SUPERVISED LEARNING (SSL)

8. Hebbian two-factor rules (Self-supervised Learning 1)

9. Two-factor rules for independent factors (Self-supervised Learning 2)

10. Learning of representations in multi-layer networks (Self-supervised Learning 3)

COMBINING RULES OF RL AND SSL

11. Learning to find a goal: a bio-plausible model with place cells and rewards (Application 2)

12. Learning by surprise and novelty: exploration and changing environments (Application 3)

13. Surprise and novelty in changing environments (Application 4)