Benchmark of Tensor Network Algorithms with Uni10
Yun-Hsuan Chou1*, Ying-Jer Kao1
1Department of Physics, National Taiwan University, Taipei, Taiwan
* Presenter:Yun-Hsuan Chou, email:hunterlj3201usa@gmail.com
Since the numerical renormalization group (NRG) came out, various numerical methods sprang up like mushrooms for determining the phase transition of many body systems and the physical properties of macroscopic systems from microscopic description. Especially, the algorithms based on the tensor network (TN) representation play significant roles in the field.

However, with the difficulty of physics models increased and the limitation of conventional machines, the TN algorithms become more and more complicated and hard to implement in the recent years.

Hence, we designed a flexible and extendable library Uni10, an object-oriented C++ library, to provide a simple API for developing TN algorithms. With Uni10, users are able to build a symmetric tensor from a collection of bonds and apply the network class to contract complicated tensor network diagrams painlessly. In this oral, I will briefly introduce some popular TN algorithms and show how to benchmark them with Uni10.


Keywords: Uni10, Tensor Network, Numerical Methods