Cognitive Models for Learning to Control Dynamic Systems
Abstract
Wald's sequential probability ratio test (SPRT) model and the equivalent discrete drift diffusion model have been widely used to explain human and animal decision making in psychophysical tasks. These models assume that observers gradually accumulate evidence from noisy inputs and make a decision when the evidence reaches a threshold. It is discovered that stochastic-resonance (SR) like behavior arises in the SPRT model when the actual input signal is significantly weaker than anticipated by the model. Analytical expressions and conditions for the SR-like behavior are found. Therefore appropriate amount of noise can improve the decision making process when the input signal is significantly weaker than anticipated. Research in adaptive SPRT demonstrates that there is an optimal nonzero and finite weight to achieve the best accuracy when the real distributions are wider than the prior distribution.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 26, 2008
- Accession Number
- ADA488027
Entities
People
- Zhong-Ping Jiang