design-bench 1.4

Design-Bench: Benchmarks for Data-Driven Offline Model-Based Optimization

Benchmarks for Data-Driven Offline Model-Based Optimization

This repository contains several design benchmarks for model-based optimization. Our hope is that a common interface and stable nomenclature will encourage future research and comparability in model-based design.

Available Tasks

Current model-based design benchmarks (circa 2020) typically vary from paper-to-paper. For example, tasks employed by biologists differ strongly from those of interest to roboticists. We provide a common interface for tasks that span a wide-range of disciplines, from materials science, to reinforcement learning. We list these tasks below.

• Biology: Protein Fluorescence: design_bench.make('GFP-v0')
• Chemistry: Molecule Activity: design_bench.make('MoleculeActivity-v0')
• Materials Science: Superconductor Critical Temperature: design_bench.make('Superconductor-v0')
• Robotics: Hopper Controller: design_bench.make('HopperController-v0')
• Robotics: Ant Morphology: design_bench.make('AntMorphology-v0')
• Robotics: DKitty Morphology: design_bench.make('DKittyMorphology-v0')

In addition, the following debugging tasks are provided.

• Debugging: Quadratic Maximization: design_bench.make('Quadratic-v0')

Setup

You can install our benchmarks with the following command.

pip install design-bench[all]

If you do not have a MuJoCo License you can install the base benchmark.

pip install design-bench

Usage

Every task inherits from the design_bench.task.Task class. This class provides access to attributes task.x and task.y that correspond to designs and labels as numpy arrays, respectively. In addition, every task implements a task.score(x) function that provides an (approximate) oracle predictor for task.y.

import design_bench
task = design_bench.make('Superconductor-v0')
x = task.x[:10]
y = task.y[:10]
oracle_y = task.score(x)