πŸ•Έ Tensor Networks β€” Structure, Compression, and Quantum Many-Body Learning

1 Overview

Tensor networks are structured factorizations of high-order tensors that make otherwise intractable computations feasible. They are central in quantum many-body physics, where they provide compact representations of states and operators with limited entanglement, and they are increasingly important in machine learning for compression, inductive bias, and scalable modeling.

Instead of storing a full tensor with exponentially many coefficients, tensor-network ansatze decompose the object into local tensors connected by virtual indices (bonds). The bond dimensions control expressivity and computational cost, creating a practical tradeoff between accuracy and efficiency.

This page is an onboarding entry point to the tensor-network toolkit used across simulation and learning: it summarizes the key ideas, points to software ecosystems, and gathers references for rigorous study.

2 Core Concepts to Master

  1. Tensor rank, index contraction, and graphical notation
  2. Matrix Product States (MPS) and Matrix Product Operators (MPO)
  3. Entanglement area laws and bond-dimension scaling
  4. Singular Value Decompositon (SVD)
  5. Core algorithms: DMRG, TEBD, and variational optimization
  6. Practical error control: truncation criteria and conditioning

3 Main Software Ecosystem

4 References

4.1 Textbooks and Lecture Notes

  • Bridgeman, J. C., & Chubb, C. T. Hand-waving and Interpretive Dance: An Introductory Course on Tensor Networks. Journal of Physics A 50, 223001 (2017).
  • Biamonte, J., & Bergholm, V. Tensor Networks in a Nutshell. arXiv:1708.00006.

4.2 Tensor Networks and Machine Learning

  • Stoudenmire, E., & Schwab, D. J. β€œSupervised Learning with Tensor Networks.” NeurIPS (2016).

4.3 Practical Learning Resources