camd is software designed to support Computational Autonomy for Materials Discovery
based on ongoing work led by the
Toyota Research Institute.
camd enables the construction of sequential learning pipelines using a set of
abstractions that include
- Agents - decision making entities which select experiments to run from pre-determined
candidate sets
- Experiments - experimental procedures which augment candidate data in a way that
facilitates further experiment selection
- Analyzers - Post-processing procedures which frame experimental results in the context
of candidate or seed datasets
In addition to these abstractions, camd provides a loop construct which executes
the sequence of hypothesize-experiment-analyze by the Agent, Experiment, and Analyzer,
respectively. Simulations of agent performance can also be conducted using
after the fact sampling of known data.