3 things the brain has to say about learning
July 29th, 2009 by Francois Rivest
- The world is dynamic, the brain is dynamic: There is no such thing called static processing. One can’t learn only from unordered pairs of inputs and outputs.
- There is no unique St-Graal rule of learning: Learning is like linguistic, it has a number of sub-problems, for which there must be more specialised solutions, and these solutions are dependent on each others.
- Hierarchy: Only when one has accumulated enough information at one level of complexity, can he raise its level of abstraction to the next level, and each level can then cover a slightly larger portion of the input space.
Computational Neuroscience (and Programming) Blog
3: By virtue of 2, learning processes themselves must be subject to evaluation and change, and I don’t see why the brain would wait for sufficient data to be gathered to subject it learning process to plasticity. The learning process’s adequacy is in the learning itself, so they should probably evolve at the same time.
Right I agree, I think what I mean is more like: You need to have learned enough about some level of complexity, before increasing the complexity of the representation… learning at multiple level of complexity may append at the same time, but higher levels may well lag on lower levels… (think about language acquisition in children for example)
But there is also parallel learning among parallel systems (cerebellum, cortex, hippocampus, basal ganglia, etc.).