CoTenN: Constrained Optimization with Tensor Networks
summary
A constraint-first DSL and compiler pipeline converts discrete quantum constrained optimization problems into tensor networks: Masq DSL expresses hard constraints, CoTenN compiler emits sparse tensor networks (including a DFA→tensor-network mapping for local constraints) that plug into DMRG-style eigensolvers. On KL and PXP benchmarks the projection-style constraint encodings cut iterations and runtime vs regularization (reported 2.05×–53.32× speedups) and reduce truncation error dramatically (up to 2.49·10^7×), demonstrating automatic, faster, and more accurate ground-state searches.
AI-written summary · about
chapters (8)
- 0:00CoTenN introduction
- 1:27Constraints in tensor networks
- 2:49Example problem and Mask DSL
- 5:57Constraint handling methods
- 7:01Compiling constraints
- 11:12Physics benchmarks performance
- 14:43Results and future work
- 15:45Q&A
related talks
01h 07mTwo Tensor Optimization Problems From Quantum Information
27mA Haskell Adiabatic DSL: Solving Classical Optimization Problems on Quantum Hardware
20mQuantum Simulation with Sparse Tensors
42mNoncommutative Polynomial Optimization under Dimension Constraints
16mqTPU: Hybrid Tensor Networks for Quantum-Classical Acceleration
59mUtility, Computational Cost, and Effective Quantum Volume of NISQ Experiments
30mQuantum Lego: Quantum Codes and their Enumerators from Tensor Networks
01h 04mConvex Optimization and Quantum Information