Variational convolutional neural networks classifiers (Q2675928): Difference between revisions

In the introduction part of the article, we find a short review on significant literature related to NISQ (noisy intermediate scale quantum) devices, quantum computers, QML (quantum machine learning), QNN (quantum neural networks), HOC (hierarchical quantum classifiers), VQTN (variational quantum tensor networks), PQC (parameterized quantum circuits), CNN (convolutional neural networks) and its quantum version. The aim of the paper here is to present a VCNN (variational convolutional neural networks) algorithm intended to solve classification problems. The algorithm utilizes parametric quantum circuits. A résumé of the contribution is: starting from a CNN architecture, convolutional and pooling layers are replaced by PQC. The second section of the paper is devoted to a detailed description and presentation of the new Algorithm 1: The VCNN algorithm, together with data encoding, circuit architecture, learning process and complexity analysis. The third section contains the results on numerical simulations. The VCNN experiments have been done on the classical MNIST and Fashion MNIST datasets using -- according to this report -- a special TensorFlow Quantum platform. Performance results on binary classification, multi-class classification, three-class classification, four-class classification and resilience to noise are shown and discussed. In the last section, some conclusions, motivations and future trends are presented.