Quantum k-medians clustering#

Code at: https://github.com/qiboteam/qibo/tree/master/examples/qclustering

Implementation of QKmedians from the paper: 2301.10780.

Before using install additional package:

  • h5py

Algorithm’s pseudocode#

pseudo

Distance calculation quantum circuit#

Distance circuit

How to run an example?#

Scripts are using qibojit as default backend.

Download dataset#

Dataset’s dimensionality is reduced by passing it through autoencoder. If you are interested more, please refer to [*].
Reduced dataset can be downloaded from Zenodo : record/7673769
Small portion of dataset in data folder:

  • latentrep_QCD_sig.h5: train dataset (QCD)

  • latentrep_QCD_sig_testclustering.h5: test dataset (QCD)

  • latentrep_RSGraviton_WW_NA_35.h5: test dataset (Signal)

Run training#

To run a training of quantum k-medians algorithm we need to provide arguments:

  • train_size (int): number of samples for training

  • read_file (str): path to the training dataset

  • seed (int): seed for consistent results in training

  • k (int): number of clusters (default = 2)

  • tolerance (float): convergence tolerance (default = 1.0e-3)

  • min_type (str): minimization type for distance to cluster search (default = 'classic')

  • nshots (int): number of shots for executing quantum circuit (default = 10000)

  • save_dir (str): path to save results

  • verbose (bool): print log messages during the training if True

  • nprint (int): print loss function value each nprint epochs if verbose is True

python train_qkmedians.py --train_size 600 --read_file 'data/latentrep_QCD_sig.h5' --k 2 --seed 123 --tolerance 1e-3 --min_type 'classic' --save_dir 'output_dir' --verbose true --nprint 1

Run evaluation#

To run an evaluation of quantum k-medians algorithm we need to provide arguments:

  • centroids_file (str): name of the file for saved centroids coordinates

  • data_qcd_file (str): name of the file for test QCD dataset

  • data_signal_file (str): name of the file for test signal dataset

  • k (int): number of clusters (default = 2)

  • test_size (int): number of test samples (default = 10000)

  • title (str): Title of ROC curve plot (default = 'Anomaly detection results')

  • results_dir (str): path to file with saved centroids

  • data_dir (str): path to file with test datasets

  • save_dir_roc (str): path to directory for saving ROC plot

  • xlabel (str): name of x-axis in ROC plot

  • ylabel (str): name of y-axis in ROC plot

python evaluate.py --centroids_file 'centroids.npy' --data_qcd_file 'latentrep_QCD_sig_testclustering.h5' --data_signal_file 'latentrep_RSGraviton_WW_NA_35.h5' --results_dir 'output_dir' --data_dir 'data' --save_dir_roc 'output_dir'

ROC curve plot#

Output of evaluation script ROC_curve