Reproduce “SeQuant Framework for Symbolic and Numerical Tensor Algebra. I. Core Capabilities” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bimal Gaudel, Robert G. Adam, Ajay Melekamburath, Conner Masteran, Nakul Teke, Azam Besharatnik, Andreas Köhn, Edward F. Valeev
SeQuant is an open-source C++ framework for tensor equations. It represents tensor networks as colored graphs to identify equivalent index arrangements and canonicalize expressions, supporting simplification and avoiding redundant Wick contractions. In synthetic tests, its canonicalization grew approximately quadratically with slots, while Butler–Portugal methods slowed on networks with equivalent tensors; this is not a general runtime guarantee.
Key insights
- Representing complex tensor networks as colored graphs allows for robust canonicalization with polynomial scaling, outperforming conventional group-theoretic methods for many tensor types.
- Incorporating topological optimizations derived from tensor network canonicalization dramatically reduces the number of terms processed by Wick's theorem, leading to orders of magnitude speedup.
- A direct interpreter-based approach with an optimized intermediate representation and common subexpression elimination can efficiently bridge symbolic manipulation and numerical execution, enhancing reusability and performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.