Learning in binary neural nets which have fuzzy reasoning
J.A.B. Tome · 2005
A binary neural model is presented, which allows for a natural way of learning and reasoning. Learning is a two phase process where connections are first established and then functions defined. An important fact is that this learning is completely unsupervised and based only on the neural activity induced by experiments. It is shown that fuzzy reasoning is naturally induced in this model without any centralized commands or explicitation. Learning may be performed in a mixed way, that is more then one rule at a time, and with any degree of uncertainty about the rules at the moment of learning.