This repository contains the the results presented in the paper A Variational Algorithm for Quantum Single Layer Perceptron submitted in the The 8th International Online & Onsite Conference on Machine Learning, Optimization, and Data Science (LOD 22).
The quantum Single Layer Perceptron (qSLP) generates an exponentially large number of parametrized linear combinations in superposition that can be learnt using quantum-classical optimization. As a consequence, the number of hidden neurons scales exponentially with the number of qubits and, thanks to the universal approximation theorem, our algorithm opens to the possibility of approximating any function on quantum computers. The repository contains also a comparison between the qSLP as a classification model against two different quantum models on two different real-world datasets usually adopted for benchmarking classical machine learning algorithms (MNIST and Iris).
- experiments: folder with the notebooks able to reproduce the experiments. It contains:
- experiments_qSLP.ipynb: file to reproduce the experiments involving the proposed model. You only need to choose the number of control qubits and the state preparation strategy. If the seed and the starting points are set it produces the results seen in the research
- experiments_qNNC.ipynb: file to reproduce the qnnc models' performance. If the seed and the starting point are set it reproduces the results obtained in the research
- experiments_QSVC.ipynb: file to reproduce the QSVC model performance. If the seed and the starting point are set it reproduces the results obtained in the research
- results: a folder containing the results of the testing:
- test_real: results obtained by the best models in a real quantum environment
- test_simulation: results obtained by the best models in a simulated environment
- training: all the models trained (qSLP, qNNC, QSVC)
- model_training.py: script to perform the training and testing on a simulated environment
- real_device_test.py: script to perform the testing of the best models on a real environment
- result_analysis.ipynb: notebook for the visualization of the performance of the models.
The code is organized in three main parts:
- Training and testing on the simulation are done through the model_triaining.py script. Run it and on completion, it writes the results on the results/training and results/test_simulation folders. To run the script simply use
python Variational_Algorithm_qSLP/model_training.py - Testing on a real device is done through the real_device_test.py script. To run it there should be an IBMQ token in the system to have access to the IBM's quantum devices. Results are stored in the folder results/test_real.
python Variational_Algorithm_qSLP/real_device_test.py - To run and test the best model of each category refer to the notebooks in the experimetns_qSLP folder. There you can choose the specific model of each category to run, obtaining the results shown in the research.
To run the code and reproduce the results of the paper, it is recommended to re-create the same testing environment following the procedure below.
Note: it's assumed Anaconda is installed
- First clone the repository
- Second, create a conda environment from scratch and install the requirements specified in the requirements.txt file:
# enter the repository cd project_folder # create an environment with desired dependencies found in the requirements.txt file conda env create pip install -r requirements.txt