Abstract
Methodologies for automated agents rely critically on decision-making algorithms with provable performance that successfully address the many tensions that arise from multiple performance objectives. The heuristics employed by humans in decision making have the advantage of both computational tractability and biological plausibility, there is great promise in leveraging them in the design of algorithms. A research paper by Reverdy, Srivastava, and Leonard discusses provable algorithms for decision making under uncertainty drawing from research on heuristics used by humans in the tradeoff between exploration and exploitation. The stochastic upper credible limit (UCL) algorithm developed in their paper connects the studied human heuristics with a decision maker that uses Bayesian inference together with the optimism in the face of uncertainty strategy, following. Other features of the algorithm reflective of human behavior include decision-making noise, the role of finite time horizon, and familiarity with the environment and its correlation structure.
Original language | English (US) |
---|---|
Article number | 6777969 |
Pages (from-to) | 572-573 |
Number of pages | 2 |
Journal | Proceedings of the IEEE |
Volume | 102 |
Issue number | 4 |
DOIs | |
State | Published - Apr 2014 |
Externally published | Yes |
All Science Journal Classification (ASJC) codes
- Electrical and Electronic Engineering