1.4 Computational Modeling as a Method to Understand Perceptual Processes

Mathematical Model

A model that uses mathematical language, concepts, and equations to closely mimic psychological and neural processes with mathematical precision.

Computational Models

Models that use powerful computers to describe and understand sensation and perception in ways that would be difficult or impossible with manual techniques.

Statistical Optimization Models

Models that use statistics from past perceptual experiences to help explain and optimize perceptual processes.

Efficient Coding Model

A model that attempts to maximize the amount of information transmitted by a perceptual system by putting less emphasis on predictable or redundant signals that provide little new information.

Maximum-Likelihood Estimator

An estimator that considers multiple sensory signals but gives more weight to the sense that has been more reliable based on past experience.

Bayesian Models

Models that use statistics from past experience to interpret the sensory information currently arriving at the senses.

Artificial Neural Networks

Also called connectionist models. Models made of layers of heavily interconnected computational units that can learn either with or without feedback.

• Supervised: learns with feedback about whether it is right or wrong.

• Unsupervised: learns without feedback about whether it is right or wrong.

Deep Neural Networks (DNNs)

Also called deep learning models. Neural networks with many layers of units and millions of connections that use modern computers to process and classify huge amounts of information into categories.