Hermes is a repository designed to be a modular compilation of tools for machine learning and autonomous experimentation in materials science domains.
This code base is under active development.
We use poetry as our main package and dependency manager. We recommend using poetry to
install hermes. To do so, clone this repo, navigate to the root directory and run poetry install.
For instructions on how to install poetry see here
Alternatively, you can run pip install . inside the root directory to install hermes. If your machine is
macOS and ARM64 (M1, M2), this is the recommended method.
To install hermes without cloning this repository, run the following command:
$ pip install git+ssh://git@github.com/cvelezrmc/hermes.git@scratchcvor to run without SSH
$ pip install git+https://<my_token>@github.com/cvelezrmc/hermes.git@scratchcvwhere <my_token> is your personal access GitHub Token.
- Consistent active learning and modeling interface aimed at enabling nonstandard analysis and acquisition policy.
- Data acquisition and wrangling with no-work FAIR backend integration
- Actual ML models and bag of materials/physics tricks.
-
Instrument Communication:
-
Basic functions for importing data from instruments and setting them up for use in modeling -
Instrument specific functions for reading data in, sending commands and the like. -
Intrinsic Data Analysis:
-
Analysis of the intrinsic properties of the data -
Examples include: data pre-processing, domain-specific data manipulation, -
clustering, dimesionallity reduction, distance measures. -
All inputs are treated as features -
Relational Data Analysis:
-
Analysis of how the inputs are related to observations of the outputs. -
Examples include: Regression, classification, physical models. -
Persistant Storage:
-
Basic functions for data storage and database design/use.
Austin McDannald
austin.mcdannald@nist.gov
National Institute of Standards and Technology
Material Measurement Laboratory
Materials Measurment Science Division
Data and AI-Driven Materials Science Group