Error-Driven Stochastic Search for Theories and Concepts
Owen Lewis, Santiago T. Pérez, Joshua B. Tenenbaum · eScholarship (California Digital Library) · 2014
Bayesian models have been strikingly successful in a wide range of domains.However, the stochastic search algorithms generally used by these models have been criticized for not capturing the error-driven nature of human learning.Here, we incorporate error-driven proposals into a stochastic search algorithm and evaluate its performance on concept and theory learning problems.Compared to a model with random proposals, we find that error-driven search requires fewer proposals and fewer evaluations against labelled data.