pen-corsica / gridsearch

Wu, C. M., Schulz, E., Speekenbrink, M., Nelson, J. D., & Meder, B. (2018). Generalization guides human exploration in vast decision spaces. Nature Human Behaviour. 2, 915–924. https://doi.org/10.1038/s41562-018-0467-4

Gridsearch

Complete code and data for replicating the results from:

Wu, C. M., Schulz, E., Speekenbrink, M., Nelson, J. D., & Meder, B. (2018). Generalization guides human exploration in vast decision spaces. Nature Human Behaviour. 2, 915–924. https://doi.org/10.1038/s41562-018-0467-4 (pdf)

An earlier version of this project (Experiment 2D and early modeling results) was released as:

Wu, C.M., Schulz, E., Speekenbrink, M., Nelson, J.D., Meder, B. (2017). Mapping the unknown: The spatially correlated multi-armed bandit. In G. Gunzelmann, A. Howes, T. Tenbrink, & E. J. Davelaar (Eds.), Proceedings of the 39th Annual Conference of the Cognitive Science Society (pp. 1357-1362). Austin, TX: Cognitive Science Society. doi: https://doi.org/10.1101/106286

Getting Started and prerequisites

The environments used in Experiments 1 and 2 were generated using samplePrior1D.py and samplePrior2D.py, which requires numpy, GPy, and Scikit-learn. Experiment 3 uses a variety of agricultural data sampled from the R package agridat (see SI for full details), which are saved in /experiment3/environments/agridat.json.

99.9% of the rest of the data analysis uses R, with code separated into /analysis1D, /analysis2D, and /analysis3. Required packages are specified at the top of each file. Model recovery code has it's own folder called /modelRecovery. Additionally, mismatch simulations are in /mismatch, where empirical simulations for each experiment are in R, while the generalized mismatch simulation is in python using the bayesian optimization library (https://github.com/jmetzen/bayesian_optimization). Mismatch simulations for Experiment 3 are in /analysis3/mismatch.R.

Experiment code is added as well, which can be found in /experiment1D, /experiment2D, and /experiment3. The code has been modified to remove MySQL functionality, which depended on PHP code that would have exposed credentials. Now experimental conditions are assigned randomly (instead of psuedo-randomly using pre-generated conditions), and experiment data doesn't get saved anymore.

Authors

See the list of contributors who participated in this project.

License

This project is licensed under the MIT License - see the LICENSE.md file for details