/badapted

Bayesian ADAPTive Experimental Design

Primary LanguagePythonMIT LicenseMIT

Baselines for DAD

This repo is forked from the main Badapted repo, and provides the implementation of the baselines used for the hyperbolic temportal discounting experiment in the paper Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Design.

Environment setup and computing infrastructure

We ran this code on a CPU-only machine with 40 CPU cores. For installation, please follow the instructions on the original repo.

Badapted and Frye baselines

Use

$ python3 badapted_hyperbolic.py

to generate 10000 rollouts using the Badapted method. Use

$ python3 frye_hyperbolic.py

to generate 10000 rollouts using the Frye method.