@article {Wu171371, author = {Charley M. Wu and Eric Schulz and Maarten Speekenbrink and Jonathan D. Nelson and Bj{\"o}rn Meder}, title = {Exploration and generalization in vast spaces}, elocation-id = {171371}, year = {2017}, doi = {10.1101/171371}, publisher = {Cold Spring Harbor Laboratory}, abstract = {Foraging for food, developing new medicines, and learning complex games are search problems with vast numbers of possible actions. Under time or resource constraints, optimal solutions are generally unobtainable. How do humans generalize and learn which actions to take when not all outcomes can be explored? We present two behavioural experiments and competitively test 27 models for predicting individual search decisions. We find that a Bayesian function learning model, combined with an optimistic sampling strategy, robustly captures how humans use generalization to guide search behaviour. Taken together, these two form a model of exploration and generalization that leads to reproducible and psychologically meaningful parameter estimates, providing novel insights into the nature of human search in vast spaces. Importantly, our modelling results and parameter estimates are recoverable, and can be used to simulate human-like performance, bridging a critical gap between human and machine learning.}, URL = {https://www.biorxiv.org/content/early/2017/08/01/171371}, eprint = {https://www.biorxiv.org/content/early/2017/08/01/171371.full.pdf}, journal = {bioRxiv} }