Neural Learning of Temporal Structures
Abstract
The goal of this proposed effort is to develop a normative, descriptive and biologically plausible mathematical framework for a fundamental capability of the human brain. More specifically, the performer will focus on human perception of uncertainty and randomness, and propose to develop a time-based mathematical treatment for neural learning and decision-making under uncertainty. They will show that, due to the fact that timeÕs arrow is irreversible, the brain has to learn in the context of time, and that the brain is capable of learning time-based structures. The scope of work would consist of three major components: ¥ The performer will provide a detailed analysis on why time is critical to the brain and its function and how current dominant theories fail to incorporate time and are therefore inadequate. This analysis will be based on both theoretical work on general stochastic process and empirical evidence on how the brain learns and reacts in uncertain situations. ¥ The performer will then attempt to develop a new mathematical treatment for neural learning that emphasizes time. We will show how the new learning system is simultaneously normative, descriptive, and biological plausible, and, more critically, allows the brain to learn something from ÒnothingÓ (in the sense that the critical temporal structures are washed out in traditional frequency-based treatments). ¥ The performer will show how the new treatment can be used to explain a range of cognitive phenomena that have proven difficult for current dominant time-less theories. Examples include the representativeness heuristic and the GamblerÕs Fallacy.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 10, 2016
- Source ID
- N000141612111
Entities
People
- Hongbin Wang
Organizations
- Office of Naval Research
- Texas A&M University System
- United States Navy