Einsum: The tensor expression language
Einsum has become the quasi-standard for tensor expressions. It is widely used in machine learning and quantum computing. We work on einsum as a compute backend for optimization, learning, and inference, providing efficient execution libraries.
Our Services
Explore our suite of research tools and platforms available on dedicated subdomains.
Research Papers
Our latest contributions to the scientific community.
Einsum Trees: An Abstraction for Optimizing the Execution of Tensor Expressions
Alexander Breuer, Mark Blacher, Max Engel, Joachim Giesen, Alexander Heinecke, Julien Klaus, Stefan Remke
The Syntax and Semantics of einsum
Maurice Wenig, P. G. Rump, Mark Blacher, Joachim Giesen
Optimizing Tensor Contraction Paths: A Greedy Algorithm Approach With Improved Cost Functions
Sheela Orgler, Mark Blacher
Model Counting and Sampling via Semiring Extensions
Andreas Goral, Joachim Giesen, Mark Blacher, Christoph Staudt, Julien Klaus
Serving MPE Queries on Tensor Networks by Computing Derivatives
Maurice Wenig, Hanno Barschel, Joachim Giesen, Andreas Goral, Mark Blacher
Einsum Benchmark: Enabling the Development of Next-Generation Tensor Execution Engines
Mark Blacher, Christoph Staudt, Julien Klaus, Maurice Wenig, Niklas Merk, Alexander Breuer, Max Engel, Soren Laue, Joachim Giesen
Improved Cut Strategy for Tensor Network Contraction Orders
Christoph Staudt, Mark Blacher, Julien Klaus, Farin Lippmann, Joachim Giesen
Compiling Tensor Expressions into Einsum
Julien Klaus, Mark Blacher, Joachim Giesen