CGSC170 Chapter Notes - Chapter 8.3: Backpropagation, Hebbian Theory, Unsupervised Learning
20 Dec 2019
School
Department
Course
Professor
Document Summary
Individual unit"s weights changes directly as a function of the inputs to and outputs from that unit. Information for changing the weight of a synaptic connection is directly available to the presynaptic axon and the postsynaptic dendrite. Neural network modelers think of it as much more biologically plausible than backpropagation. And a way of spreading an error signal back through the network. No fixed target for each output unit. Classified set of inputs in such a way that each output unit fires in response to a particular set of input patterns. Require detecting similarities between different input patterns. Ex- they have been used for modeling visual pattern recognition. Visual recognition is able to see the same object from multiple angles and perspectives. There are several competitive network models of this type of position-invariant object recognition. No connections between units in a single layer. Allow the output units to compete with one another.
