COGS 100 Lecture Notes - Lecture 9: Distributed Computing, Deep Learning, Step Function
22 Apr 2015
School
Department
Course
Professor
Document Summary
Maybe computer is useful model for mind. Anns (connectionism, parallel distributed processing, pdp, deep learning) Emphasis on patterns of activation across system of nodes and links rather than localized representation and algorithmic processing. Activation level of node x = sum of (input node values) times (weight) for all input links. Step function approach: like a neuron, node x fires or not depending on whether these summed activation levels reach a threshold. More complex approaches allow node x to fire at different strengths, depending on how much input it. In serial processing, computation is performed one step at a time. The brain, however, computes in a massively parallel and distributed way, with different operations ongoing at any given time. Systematic study of language phonetics & phonology (sounds) Morphology (meaning) syntax (structure) semantics (meaning) pragmatics (context) Phoneme smallest units of sound that make up a language. Each language has a distinctive set of phonemes.
