Part 1: Object recognition Part 2: Computational modelling
Object recognition
Overview
Apperceptive and associative agnosia
Warrington's two-stage model of object recognition
Warrington's Unusual Views and Shadows Tests for apperceptive agnosia
Right hemisphere lesion Unusual Views Test
Right hemisphere lesion (cont'd) Shadows Test
Associative agnosia: semantic categorization is impaired
Hemi-neglect
Bisect all the lines..., a test for hemineglect
Different visual stimulus arrays
Evidence for contralateral inhibition
Evidence for ipsilateral exitation
Neglect distributed in objects
Neglect in imaging
The code of the brain
Types of neural representations
Extremely localized coding leads to the grandmother cell
Sparse coding
Desimone's study of V4* neurons
Neurons in IT show evidence of `short-term memory' for events
PPT Slide
Reduced IT response and memory
Novelty filtering
What and where streams
Where stream
Neuron in posterior parietal cortex
What stream
A neuron in inferior temporal cortex (IT)
What is known about what is located in the brain?
PET data corroborate the lesion data
Computational modelling
Neural networks
Neural networks abstract from the details of real neurons
Artificial `neuron'
Illustration of a neural network
Much of perception is dealing with ambiguity
Many interpretations are processed in parallel
The final interpretation must satisfy many constraints
i. Only one word can occur at a given position
ii. Only one letter can occur at a given position
iii. A letter-on-a-position activates a word
iv. A feature-on-a-position activates a letter
Recognition of a letter is a process of constraint satisfaction
How to `program' neural networks?
Neural networks and David Marr's model (1969)
Hebb (1949)
The Hebb rule is found with long term potentiation (LTP) in the hippocampus
Willshaw networks
Example of a simple heteroassociative memory of the Willshaw type
Example of pattern retrieval
Example of successful pattern completion using a subpattern
Example graceful degradation: small lesions have small effects
Error-correcting learning
The Perceptron by Frank Rosenblatt (1958, 1962)
Very simple example
Learning problem to be solved
Answer
Perceptron algorithm in words
Error-backpropagation
Error-backpropagation by Rumelhart, Hinton, and Williams
Backpropagation algorithm in rules
Despite its popularity backpropagation has some disadvantages
Good points
Email: jaap@murre.com
Home Page: http://www.neuromod.org/courses/local.html
Other information: This is a lecture in a course on Clinical Neuropsychology.
Download presentation source
University of Amsterdam
