Quantum Expectation-Maximization for Gaussian mixture models

Authors: Iordanis Kerenidis, Alessandro Luongo, Anupam Prakash

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this work, we tested our algorithm on a non-trivial dataset for the problem of speaker recognition on the Vox Forge dataset (Voxforge.org). The experiment aimed to gauge the range of the parameters that affects the runtime, and test if the error introduced in the quantum procedures still allow the models to be useful. From the experiments reported in Section 5, we believe that datasets where the number of samples is very large might be processed faster on a quantum computer.
Researcher Affiliation Collaboration 1Institute de Recherche en Informatique Fondamental Paris, France 2QC Ware Corp. Palo Alto, California 3Atos Quantum Lab Les Clayes sous Bois, France.
Pseudocode Yes Algorithm 1 QEM for GMM
Open Source Code No The paper does not provide any specific repository link or explicit statement about the release of its source code.
Open Datasets Yes In this work, we tested our algorithm on a non-trivial dataset for the problem of speaker recognition on the Vox Forge dataset (Voxforge.org).
Dataset Splits No The paper mentions 'training set' and 'test set' but does not provide specific details on dataset splits (e.g., percentages for train/validation/test, cross-validation setup, or sample counts for each split).
Hardware Specification No The experiment has been carried in form of classical simulation on a laptop computer.
Software Dependencies No The paper mentions 'scikit-learn' but does not provide specific version numbers for any software components, programming languages, or libraries used in the experiments.
Experiment Setup Yes For values of δθ = 0.038, δµ = 0.5, we correctly classified 98.7% utterances. The value of the threshold of the likelihood ϵτ is usually (for instance in scikit-learn (Pedregosa et al., 2011) ) chosen to be 10 3.