|
"
Action for learning in non-symbolic systems
"
Download: |
This paper claims that neural networks should learn in a progressive way, starting with easy tasks and continuing with more difficult ones.
A review of the subject reveals that only a tiny mount of research was done in that direction.
We thus performed two basic experimental studies, showing that progressive learning could effectively facilitate learning.
Beyond those "toy" experiments, how progressive learning is performed in complex systems like the human cognitive system? We found that the psychological theories of automatization constitutes an interesting model of progressive learning. These theories assume that humans first process information attentionaly selectively filtering information and then automates processing. A prospective model is proposed in that sense, at the end of the paper. |
(( TOPICS
Progressive learning (( LABS Lifia |