Applying Variational Circuits in Deep Learning Architectures for Improving Discriminative Power of Speaker Identification Embeddings

  • Рафаэль Бланксон

Student thesis: Master's Thesis


Recently, the advancement in quantum technologies has had a massive impact on the development of quantum algorithms on near-term quantum devices. Variational circuits, a combination of both quantum and classical algorithms, have been very useful in these advancements on near-term quantum devices. Despite these advances, most quantum applications in machine learning
(deep learning) especially in transfer learning have been proof-of-concept in the qubit system and very little in the continuousvariable space but no or little application to audio data.
This study applies variational circuits to practical real-life speaker classification data for the first time in the continuous-variable system to investigate whether the speaker embeddings can be improved by applying quantum models. In separate experiments, the quantum model was combined with a simple convolutional neural network and ResNet18 model respectively and the results was compared to the classical ResNet18 model applied on the same speaker dataset. The simple convolutional with quantum model outperformed the ResNet18 quantum model significantly but were worse compared to the classical ResNet18 model. The use of the mixup algorithm significantly improved the performance of the quantum models.
Further investigation is needed to model-specific problems that classical models cannot solve and to show a quantum advantage.
Date of Award15 Jun 2021
Original languageEnglish
Awarding Institution
  • Novosibirsk State University
SupervisorЕвгений Николаевич Павловский (Supervisor)

Cite this