Last minute repetition Flashcards
Connectionism
- Interconnected networks of simple units which work parallel to each other, connected by links
- Consists of three layers: input, hidden and output units
Constraint satisfaction
- Helps networks find stable patterns of activation that satisfy multiple interacting constraints
- Key to parallel distributed processing (PDP) models, where units adjust to reach a stable state
Differential Hebbian Learning
A variant of Hebbian learning where weight changes depend on the temporal difference of pre- and postsynaptic activity.
Emotional Contagion
Mechanism in which people aquire the emotions of others just by seeing their expression
Can be linked to pattern association
Types of human errors
Slips and lapse: Errors committed in familiar situations, right plan but wrong execution
Mistakes: Errors committed in unfamiliar situations, choosing the wrong plan
Simulated Annealing
Trying to escape a local minimum by adding noise, raising temperature, increasing energy and gradually cooling down to reach global minimum
SNIF-ACT
Extension of ACT-R model, designed to mimic human psychology
Information foraging theory
When humans search for information they try to maximize useful information while minimizing search effort. They look for information scents (utility) which determines whether or not something is worth pursuing.
The application of that is SNIF-ACT
Machine Learning
Ability of computers to learn without being explicitly told what to learn (self-initiated)
Stigmergy
Indirect communication by altering the state of the environment in a way that will affect the behavior of others
Can be connected to swarm intelligence
Neo-Hebbian Learning
Restatement of hebbian learning, solving the provlem of constantly increasing weights caused by too many associations
Turing machine
Mathematical model which can use a predetermined set of rules to determine a result from a set of input
Can be connected to ACT-R
How does autoassociation work in the hippocampus?
- Dentate gyrus: Pattern seperation (helps transform similar inputs into distinct representations)
- CA3: Autoassociation (Pattern completion)
- CA1: Relaying processed information