Efficient metadata management is a critical component of every computational study. The ability to robustly associate data with parameters for model parameter exploration, hyperparameter
tuning, or provenance documentation is both crucial and challenging. This talk showcases the signac framework (www.signac.io), a lean, open-source Python package designed for data and workflow management. The framework’s flexible data model allows easy adaptation into preexisting file-based workflows while also providing critical database functionality to filter, search, and group data by metadata. The first part of the talk will serve as a brief introduction to challenges commonly encountered in typical computational workflows. The second part will be a live demonstration of how we can use signac to set up and execute a computational study that avoids these pitfalls. We will then touch on how signac integrates into the larger Python ecosystem, including its use in Jupyter notebooks and in conjunction with pandas data frames. Attendees are encouraged to bring laptops to optionally follow along during the talk and are invited to stay afterwards for assistance with setting up signac for their own use.
Carl Simon Adorf is a Ph.D. candidate at the University of Michigan and the lead developer and maintainer of the core signac and the signac-flow package. His research interests include scientific computing best practices and the optimization of particle interaction models for the
self-assembly of complex materials.
Vyas Ramasubramani is a 3rd year Ph.D. candidate at the University of Michigan. He works in Sharon Glotzer’s research group, where he conducts research focused on the role of various solvent related effects on self-assembly. He is currently studying coarse-grained models of
proteins and how depletion and solvent effects can be efficiently modeled to better understand protein crystallization behavior.
Bradley Dice developed the signac-dashboard package to provide a complementary data visualization interface for the signac framework. He is a PhD student in Physics and Scientific Computing, working in the Glotzer lab at the University of Michigan. His current research uses
machine learning techniques to understand the dynamics of crystallization in colloidal materials.