QAS-Bench: Rethinking Quantum Architecture Search and A Benchmark
Authors: Xudong Lu, Kaisen Pan, Ge Yan, Jiaming Shan, Wenjie Wu, Junchi Yan
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper, beyond a particular domain, we formulate the QAS problem into two basic (and relatively even ideal) tasks: i) arbitrary quantum circuit (QC) regeneration given a target QC; ii) approximating an arbitrary unitary (oracle). The latter can be connected to the setting of various quantum machine learning tasks and other QAS applications. Based on these two tasks, we generate a public QAS benchmark including 900 random QCs and 400 random unitary matrices which is still missing in the literature. We evaluate six baseline algorithms including brute force search, simulated annealing, genetic algorithm, reinforcement learning, hybrid algorithm, and differentiable algorithm as part of our benchmark. |
| Researcher Affiliation | Academia | Xudong Lu 1 Kaisen Pan 1 Ge Yan 1 Jiaming Shan 1 Wenjie Wu 1 Junchi Yan 1 1Mo E Key Lab of AI, Shanghai Jiao Tong University, Shanghai, China. Correspondence to: Junchi Yan <yanjunchi@sjtu.edu.cn>. |
| Pseudocode | Yes | Algorithm 1 Bidirectional Brute Force Search for QC Regeneration |
| Open Source Code | Yes | Data and code are available at https://github. com/Lucky-Lance/QAS-Bench. |
| Open Datasets | Yes | For the above two tasks, by random generation with physically meaningful post-processing, we manage to provide a benchmark dataset containing 900 QCs for QC Regeneration and 400 unitary matrices for Unitary Approximation, respectively. [...] As for the train set D, we dismiss the basis states (Eq. 3), and the sampling process of the input quantum states follows Eq. 5 and Eq. 7. |
| Dataset Splits | No | The paper mentions 'train set D' and 'test set V', but it does not explicitly define or refer to a 'validation set' for model tuning or hyperparameter selection. |
| Hardware Specification | Yes | All experiments were carried out on a workstation with Intel(R) Xeon(R) Platinum 8276 CPU and 4 NVIDIA A100PCIE-40G GPUs. |
| Software Dependencies | No | The paper states 'all codes are written in Python' and 'implement baseline algorithms based on a revised version of simulator Torch Quantum (Wang et al., 2022a) with a better support for Py Torch'. While Python and PyTorch are mentioned, specific version numbers for these or other libraries like PennyLane are not provided. |
| Experiment Setup | Yes | For brute force search, we search from 1 gate, then iteratively add the gate number until the circuit is found or the time limit is exceeded. For other algorithms, we provide two ways of setting the layer number. The first one is fixing a layer number for each algorithm. Another is to set a layer number that grows linearly with the number of qubits for each algorithm. In our setting, for a n-qubit circuit, we set the initial layer number to 5n. [...] The original RL algorithm for VQE employs a DDQN (Mnih et al., 2013) with ϵ greedy policy and an Adam optimizer. |