CGSC170 Lecture Notes - Lecture 8: Connectionism, Outline Of Object Recognition
20 Dec 2019
School
Department
Course
Professor
Document Summary
Abstract mathematical tools for modeling cognitive processes. Abstract from biological details of neural functioning. Instead of capture general principles of how the brain works. A connectionist networks or artificial neural network. It is beyond replication, but it comes up with a model and a general principle. Concerns about rule based and serial models of simple cognitive abilities. People were not satisfied with simple things like turing machine. Worries about biological plausibility of physical symbol systems. There was a lack of a bridge between levels. Processing in physical symbol systems step by step. Many complicated information processing tasks can be performed in a few hundred milliseconds. But the average time it takes a neuron to generate an action potential is 3-5ms. Not very many interesting turing machine programs that can be run in 100 steps or so. If we had just a serial process where only one thing happened at once, we would have a constraint.
