ALS truncation for 3-site MPS#381
Draft
Yue-Zhengyuan wants to merge 13 commits into
Draft
Conversation
Codecov Report❌ Patch coverage is
🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR adds ALS truncation for 3-site MPS, which will find its use in FU with NNN gates.
For the first and the last (third) site, the gate acts on the rank-(2, 1) reduced bond tensor. At the middle (second) site, the gate acts on the full state tensor. Therefore, in NTU/FU applications, the MPS to be truncated looks like (for iPEPS)
=implies "thickened" bond to be truncated. The dimension of left leg of a, and right leg of b isWe then directly perform ALS sweeps over a, m, b (used in PhysRevB.97.174408). Note that truncated$D^4 d$ for iPEPS, or $D^4 d^2$ for iPEPO. In comparison, truncated $D^2 d^2$ . Therefore, we solve
mhas a large sizeaandbonly have sizea,bmore frequently.I wanted to add another approach. But I realized some subtleties and decide not to include it to make this PR more focused.
Details
To avoid directly solving$D^2 d^2$ :
m, I implement a second approach following PhysRevB.87.115144, but with some slight improvements. Here, we first find the projectors (p0, ..., p4) on each bond, each only having sizeThey can be initialized with (truncated) Vidal gauge. Then, we perform ALS sweeps over (a p0), p1, p2 and (p3 b). In tests with random tensors, I find it works slightly better than factoring out bond tensors with QR/LQ and optimizing the bond tensors. But the improved cost function is always higher than the first approach, which can be seen in the test
test/bondenv/3site_truncate.jl. It is also known to converge slowly, first shown in PhysRevB.97.174408 (Fig. 6b).