Error-Driven Stochastic Search for Theories and Concepts


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.

3 Figures and Tables

Cite this paper

@inproceedings{Lewis2014ErrorDrivenSS, title={Error-Driven Stochastic Search for Theories and Concepts}, author={Owen Lewis and Santiago Perez and Joshua B. Tenenbaum}, booktitle={CogSci}, year={2014} }