Qubit Routing Using Graph Neural Network Aided Monte Carlo Tree Search
Authors: Animesh Sinha, Utkarsh Azad, Harjinder Singh9935-9943
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The experiments were repeated 10 times on each circuit size, and final results were aggregated over this repetition. |
| Researcher Affiliation | Academia | Animesh Sinha , Utkarsh Azad , Harjinder Singh Center for Computational Natural Sciences and Bioinformatics, International Institute of Information Technology, Hyderabad. Center for Quantum Science and Technology, International Institute of Information Technology, Hyderabad. {animesh.sinha, utkarsh.azad}@research.iiit.ac.in, laltu@iiit.ac.in |
| Pseudocode | No | The paper describes the method in text and uses a flowchart (Figure 3) for the MCTS process, but does not provide explicit pseudocode blocks or algorithms labeled as 'Pseudocode' or 'Algorithm'. |
| Open Source Code | Yes | Finally, we provide a simple python package containing the implementation of QRoute, together with an easy interface for trying out different neural net architectures, combining algorithms, reward structures, etc. |
| Open Datasets | Yes | Next we test on the set of all circuits which use 100 or less gates from the IBM-Q realistic quantum circuit dataset used by Zulehner, Paler, and Wille (2018). |
| Dataset Splits | No | The paper refers to testing on 'random circuits' and 'IBM-Q realistic quantum circuit dataset', but does not provide specific details on how these datasets were split into training, validation, and test sets for their model, or mention cross-validation. |
| Hardware Specification | Yes | when tested on a personal machine with an i3 processor (3.7 GHz) and no GPU acceleration. |
| Software Dependencies | No | The paper mentions 'a simple python package' for QRoute and compares it against 'Qiskit' and 'Cirq', but does not specify version numbers for Python, these frameworks, or any other software dependencies. |
| Experiment Setup | No | The paper describes the components of the MCTS and GNN architecture (e.g., UCT, Dirichlet noise, edge-convolution block, Swish activations), but it does not provide specific hyperparameter values for training the neural network (e.g., learning rate, batch size, number of epochs) or specific parameters for the MCTS search process. |