Current browse context:
cond-mat.stat-mech
Change to browse by:
References & Citations
Condensed Matter > Statistical Mechanics
Title: Tensor Renormalization Group with Randomized Singular Value Decomposition
(Submitted on 5 Dec 2017 (this version), latest version 22 Mar 2018 (v2))
Abstract: A new algorithm of tensor renormalization group is proposed based on randomized algorithm for singular value decomposition. The new algorithm is applicable to a broad range of two-dimensional classical models. In the case of the square lattice, its computational complexity and memory usage are proportional to the fifth and the third power of the bond dimension, respectively, whereas those of the conventional implementation are the sixth and the fourth power. The oversampling parameter larger than the bond dimension is sufficient to reproduce the same result as full singular value decomposition even at the critical point of two-dimensional Ising model.
Submission history
From: Satoshi Morita [view email][v1] Tue, 5 Dec 2017 03:05:03 GMT (460kb)
[v2] Thu, 22 Mar 2018 09:46:13 GMT (1800kb,D)
Link back to: arXiv, form interface, contact.