Hi,
are there Plans to include augmented Tree Tensor Networks in the ITensor Networks library at some point?
Might there (I see how this is an anoying question) already be an approximate time frame if so?
Could you provide a link or reference for what an “augmented tree tensor network” is versus a more general tree tensor network?
The broader answer to your question is that ITensorNetworks already supports arbitrary tensor networks on any graph whatsover, and for graphs without loops i.e. trees it supports a wide range of DMRG-like algorithms (DMRG, TDVP, and so on). However, our designs for this package are rapidly evolving so a lot of it is in kind of an “alpha” level right now.
https://arxiv.org/pdf/2507.21236
It is not so much a matter of the underlying graph but of the Disentanglers
I think we probably wouldn’t support that directly in the ITensorNetworks.jl, since it is a slightly specialized ansatz, though as Miles said we are developing tools for general tree tensor networks and general graph tensor networks, so I think the tools we are adding would make it relatively easy to implement augmented tree tensor networks (say as an add-on package) by combining tools we are adding for applying gates on general graphs with the tree tensor network functionality we are working on. My guess is that part of the challenge of implementing that ansatz is handling how the Hamiltonian terms get transformed and renormalized by the disentanglers and tree tensor network layers, I think the functionality we are working on would help with that.
1 Like
Thank you