COGS 100 Lecture Notes - Lecture 18: Problem Solving, Necker Cube, Connectionism
Document Summary
Cogs 100 - introduction to cognitive science - lecture 18: connections. Problem solving is constraint satisfaction something that happens or prevents something from happening. Positive constraints: go together excite win a lottery. Both local and distributed representations can be used for parallel constraint satisfaction. Human interpretations of the necker cube can be modeled using a connectionist network. Each node represents a corner and the edges represent the connections between units with weights that represent compatibilities and incompatibilities between interpretations. Parallel constraint satisfaction converges or settles down to one of the two interpretations where a subset of nodes collectively activate and deactivate the other to represent a coherent interpretation. In local representation, each node may represent one concept or proposition. Activation of a node can be interpreted as whether a concept is applicable or the truth value of a proposition (true or false) Links can be one way (feedforward or feedback) or both ways (symmetric) network.