This repository contains the code for paper, Recognizing Good Variational Quantum Circuits with Monte Carlo Tree Search.
Dependencies we use in experiments:
numpy == 1.21.5
scikit-learn == 1.1.1
pytorch == 2.0.1
pennylane == 0.30.0
Different package versions may produce inconsistent results.
To reproduce the results of the discovered hybrid fusion network which achieves MAE of 1.138 on CMU-MOSI dataset, run schemes.py in the folder best_vqc.
The file Multibench.py contains the benchmark from MultiBench example, where the dataset is changed to ours for fair comparison.
To start a search for hybrid architectures for multimodal task of sentiment analysis, simply run MCTS.py.