Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download MIT Department of Brain and Cognitive Sciences Instructor: Professor Sebastian Seung
Document related concepts
Neural correlates of consciousness wikipedia , lookup
Cognitive development wikipedia , lookup
Catastrophic interference wikipedia , lookup
Executive functions wikipedia , lookup
Holonomic brain theory wikipedia , lookup
Neurophilosophy wikipedia , lookup
Embodied cognitive science wikipedia , lookup
Neuroethology wikipedia , lookup
Neuropsychopharmacology wikipedia , lookup
Social network (sociolinguistics) wikipedia , lookup
Psychophysics wikipedia , lookup
Stephen Grossberg wikipedia , lookup
Convolutional neural network wikipedia , lookup
Cognitive neuroscience wikipedia , lookup
Optogenetics wikipedia , lookup
Central pattern generator wikipedia , lookup
Nervous system network models wikipedia , lookup
Neuroeconomics wikipedia , lookup
Recurrent neural network wikipedia , lookup
Transcript
MIT Department of Brain and Cognitive Sciences 9.641J, Spring 2005 - Introduction to Neural Networks Instructor: Professor Sebastian Seung Lateral inhibition Truth 1: Lateral inhibition amplifies differences relative to commonalities. Truth 2: Lateral inhibition regulates response selectivity by setting a dynamic threshold Truths 1 and 2 are related: The ratio between differential gain and common gain is dynamic in a nonlinear network. Rectification vs. thresholding All-to-all inhibition Consider α<1 for now. Increasing α or β enhances selectivity Without nonlinearity Response of a linear network Piecewise linear behavior A dynamic threshold All neurons active Conditional winner-take-all The active set is scale invariant Gain depends on the active set Nonlinear amplifier • Unique steady state • State-dependent gain There is a unique output
