ITensor has a paper! We lay out the philosophy behind ITensor and give lots of examples of the key parts of the interface (mostly in Julia). We also discuss some advanced features, NDTensors.jl, and some future plans.
https://t.co/Ih7wEJ2UBk
Example model 1D system showing a beyond-group symmetry studied with ITensor DMRG.
"Non-invertible symmetry-protected topological order in a group-based cluster state" Christopher Fechisin, Nathanan Tantivasadakarn, Victor V. Albert.
https://t.co/bzwoYnvswG
@aavishkar_patel Regarding alternatives to SVD, in some settings such as TEBD, QR can be used and is fundamentally cheaper and much more highly optimized for GPU: https://t.co/9Ep0xOuKNL
Our new ITensor GPU backend system is now live in the latest Julia version, building on hard work by Karl Pierce and Matt Fishman, using the elegant new package extension system. See Matt's post about the initial release here:
https://t.co/90FzedAG0k
@aavishkar_patel Regarding bond dimension, GPU memory can be a serious limitation. We are looking into features like unified memory to help. Also moving parts of tensors between CPU and GPU or across multiple GPU's.
Beautiful agreement between numerics (DMRG based on ITensor) and Ginzburg-Landau theory calculations of coexistence of CDW and superconductivity in t-t'-U Hubbard model.
https://t.co/SyqbC8r4ej
Implemented from scratch the DMRG algorithm with @ITensorLib, which can be a helpful example on the basics of ITensors and to understand the details of DMRG's implementation.
Code:
https://t.co/YCEIpe2zKe
"Quantics tensor trains" (amplitude encoded functions represented as MPS) to represent Green's functions. Like using a tiny quantum computer to represent correlation functions of quantum systems (implemented with ITensor) https://t.co/XakPgp0811
Why do some parts of ITensors.jl have type instabilities and use dynamic dispatch, and does this hurt performance? A good answer by Matt Fishman to this question: https://t.co/wGhjKAP1Yc
Our DMRG study of the quantum critical point in 1D water is published @PhysRevLett. This work wouldn't be possible without the incredible @ITensorLib software. Also, don't miss the fantastic Viewpoint by Marivi Fernández-Serra
@Perimeter@UWaterloo
https://t.co/mDRR9PIdHN
Compression of Green and vertex functions in real and momentum space as MPS and MPO tensor networks, for solving Dyson equation and other quantum applications (partially implemented with ITensors.jl):
@onur_danaci@drpeterrohde@RichCompNeuro@JuliaLanguage ITensor is written 100% in Julia, so doesn't call any C++ or Fortran (except BLAS which every language calls, including C++). Packages like Zygote and Flux are also all Julia.
ITensor used to benchmark a state-of-art quantum simulator (Quantinuum trapped ion system used to simulate the quantum dynamics of a Floquet contact process) https://t.co/lcKUZrl6at