Tensor networks
TensNet: Tensor Networks for 2D Flow Simulation
A tensor network library using MPS and MPO for memory-efficient simulation of 2D incompressible flow.
Accepted · ICSSCC 2026
Interactive abstract
Play with the idea
This interactive diagram needs JavaScript. The schematic and write-up below explain the same idea.
Layman view
In plain words
Draft — to be verified. Written in plain language from the paper and project notes on this site.
Simulating how a fluid flows, for example air around a pillar, means storing the velocity at every point of a fine grid, which quickly needs gigabytes of memory.
TensNet stores those flow fields in a compressed form borrowed from quantum physics: tensor networks, specifically matrix product states. One setting, the bond dimension, controls how much detail is kept; larger means more accurate and more memory.
In the accepted manuscript, the compressed simulations stay close to a standard full-grid solver while using megabytes instead of gigabytes on the test cases.
Expert view
The work, step by step
-
01
Problem
Computational fluid dynamics simulations demand enormous memory for storing high-resolution velocity fields, which is a barrier to practical simulations on resource-constrained systems.
-
02
Approach
TensNet represents flow fields as Matrix Product States and operators as Matrix Product Operators, compressing them by SVD truncation with adaptive bond dimensions. The two-site DMRG algorithm solves the implicit Poisson equations in the pressure-correction scheme, and cross-approximation represents immersed boundaries and obstacle geometries as MPS. Benchmarks: a cylinder in cross-flow, a square obstacle and a 3×2 array of cylinders, compared against a dense finite-difference solution on the same grid.
-
03
Results
Reported in the accepted manuscript (ICSSCC 2026, in processing).
Setup: Intel i7 workstation with 16 GB RAM and no GPU; 256 × 64 grid, Δt = 10⁻³, ν = 0.01; accuracy measured against a dense finite-difference solution on the same grid.
At D_max = 30 the L2 error is below 10⁻² and memory falls from about 8 GB to 5.8 MB, a reduction of roughly 1.4 × 10³. The L2 velocity error falls roughly exponentially with D_max at Re = 15, 30, 100 and 500; higher Re needs a moderately larger D_max.
Scalability: on a 1024 × 256 grid with D_max = 50, TensNet uses about 35 MB; the manuscript estimates about 2 TB for a dense finite-difference solver.
Memory vs accuracy, cylinder in cross-flow (t = 1.0, 100 time steps) D_max Memory L2 error Time 10 2.3 MB 0.087 45 s 20 4.1 MB 0.032 68 s 30 5.8 MB 0.008 92 s 50 8.2 MB 0.002 145 s Dense reference 8,000 MB — 3,600 s Cylinder flow statistics, TensNet at D_max = 30 vs literature Re Cd (TensNet) Cd (reference) St (TensNet) St (reference) 15 2.41 2.35–2.40 – – 30 1.68 1.62–1.68 0.198 0.190–0.200 100 1.32 1.25–1.35 0.166 0.160–0.170 500 1.14 1.06–1.20 0.211 0.205–0.220 Computational cost, 256 × 64 grid, 100 steps Solver Memory Runtime TensNet (D_max = 30) 5.8 MB 92 s Dense finite-difference 8,000 MB 3,600 s GPU-optimised finite-difference 6,000 MB 450 s -
04
Status
Accepted at ICSSCC 2026; in processing there.
Schematic
How it works
Contribution
My role
TensNet was originally developed by N. Khan and S. S. Pandey. I carried out the testing, code troubleshooting, benchmarking and visualisation, and prepared the manuscript.
- Library: MPS/MPO, DMRGN. Khan, S. S. Pandey
- Testing and troubleshootingMe (my part)
- BenchmarkingMe (my part)
- VisualisationMe (my part)
- ManuscriptMe (my part)
Publication
The paper
2026 · Conference paper · Accepted, in processing
TensNet: Memory-Efficient Tensor Network Methods for 2D Incompressible Flow Simulation
S. Siyanwal, N. Khan, S. S. Pandey, K. Y. Islur, K. Dave, S. Chhipa
ICSSCC 2026
Abstract
TensNet is a library implementing tensor network methods, using Matrix Product States (MPS) and Matrix Product Operators (MPO), for simulating the 2D incompressible Navier–Stokes equations. It is designed to reduce memory use compared with standard grid-based solvers, keeping accuracy through adaptive bond-dimension truncation; it uses the DMRG algorithm to solve the implicit pressure-correction equations and cross-approximation to handle complex immersed boundaries.