arXiv:2610.31554
ArcTN: Efficient Tensor-Network Contraction Planning for Quantum Circuit Simulation
Zhengyi Liu, Li Liu, Songtao Liu, Xianglong Liu, Xiangzhen Zhou, Mingyu Huang, Dingchao Gao, Shenggang Ying, Bolong Chen, Ji Guan
Abstract
Tensor-network simulation of quantum circuits depends on contraction orders with low computational cost and small intermediate tensors, found without excessive planning time. We present ArcTN, a contraction-order optimization framework with two search presets. Light keeps planning overhead low, and Heavy spends more effort on refinement. Both combine randomized-greedy construction, subtree reconfiguration, and parallel tempering with stage-specific stopping rules. ArcTN also provides Fixed and Dynamic slicing to meet tensor-size limits and executes the returned orders. On the 30 QC32 quantum-circuit networks completed by all eight compared configurations, Light reduces FLOPs by 61.32% and plans 1.62$\times$ faster than Cotengra Auto/CMA-ES, and Heavy reduces FLOPs by 43.83% and plans 7.21$\times$ faster than Cotengra AutoHQ/Optuna. Heavy also needs 11.85% fewer FLOPs than OMECO TreeSA-8 and plans 1.33$\times$ faster. Dynamic slicing meets the size limits in all 90 runs on 18 scalar-output networks and reduces total sliced FLOPs by 31.99% relative to Fixed slicing, at the cost of longer planning. With a common NumPy executor, Heavy has the lowest combined planning and execution time among five configurations on the runs that all of them completed. On 672 networks built from 320 MQT Bench circuits, Heavy obtains the lowest geometric-mean FLOPs on the 142 large networks while planning 3.1$\times$ and 31$\times$ faster than Cotengra AutoHQ and Hyper128.
AI Summary
Tensor-network circuit simulation requires contraction orders that balance arithmetic cost, intermediate size, and planning overhead. ArcTN is a Rust tool with Python interfaces that coordinates existing search techniques: cross-entropy-adapted random-greedy construction, subtree reconfiguration, and parallel tempering. Light and Heavy presets allocate different refinement effort and use stage-specific stopping rules. Fixed slicing preserves the order; Dynamic slicing alternates slice selection with local order refinement, minimizing modeled work across slices subject to a per-result size limit. Plans support numerical execution and reuse. Evaluation compares against Cotengra, OMECO, and Google TNCO. On 30 QC32 instances completed by all eight configurations, Light uses 61.32% fewer modeled FLOPs and plans $1.62\times$ faster than Cotengra Auto; Heavy uses 43.83% fewer FLOPs and plans $7.21\times$ faster than AutoHQ. With a shared NumPy executor, Heavy has the lowest geometric-mean planning-plus-execution time on 134 commonly completed runs: 67.80 seconds versus 146.34 for AutoHQ. Dynamic slicing reduces total modeled sliced FLOPs relative to Fixed by 31.99% for scalar outputs and 27.04% for tensor outputs, with approximately 17% longer planning. On 142 large MQT Bench networks, Heavy achieves the lowest geometric-mean FLOPs while planning $3.1\times$ faster than AutoHQ. Results depend on workload and configuration: Cotengra achieves lower aggregate FLOPs on MQT expectation networks. Sliced plans are validated but not numerically executed in the slicing comparison; their costs exclude accumulation and output assembly. Tensor-size limits do not bound total process memory. The MQT evaluation excludes networks with at least 5,000 tensors and uses one run per configuration and network.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 29 pages, 4 figures, 15 tables. Code and benchmarks: https://github.com/Quill-ArcLight/ArcTN. v2: equal-contribution and corresponding-author notes added, author order updated, Figures 1 and 2 revised