Quantum-Enhanced Feature Maps for Improved Quantum Kernels and Advanced Quantum Machine Learning Application

Quantum machine learning has become increasingly significant in recent times, leveraging quantum-enhanced feature maps accessible only on quantum computers to potentially achieve quantum advantage (Liu, Y., Arunachalam, S., Temme K., 2021). Through this work, we explore the construction and analysis of quantum feature maps for developing enhanced quantum classifiers using the quantum kernel method, extending the application of quantum computers to machine learning (Havlíček,V., Córcoles, A.D., Temme, K. et al., 2019). We have developed enhanced quantum kernels based on six different quantum feature maps and demonstrated proof of concept with three different 2-dimensional datasets for binary classification tasks (Suzuki, Y.,Yano, H., Gao, Q. et al., 2020). Our work aims to showcase the significance of quantum feature maps and the role of hyperparameters in developing quantum support vector classifiers(QSVC) for various benchmark datasets. We have conducted a comprehensive analysis of hyperparameters, including quantum gates and rotation factors, to demonstrate their influence on data distributions, thereby offering enhanced analytical quantum advantages. Furthermore, we demonstrate the applicability and advantages of QSVC when exposed to real-world datasets. We have applied the method to brain data, which is highly complex and multivariable, containing spatio-temporal information. Spiking neural networks are employed to extract features from the EEG data, and the resulting features are utilized to develop the quantum classifier (Kasabov, 2014). The methodology yields promising results compared to various other classical machine learning models; thereby providing the quantum advantage. Finally, we claim to uniquely represent the hybrid approach for bridging neuromorphic and quantum computing, capable of providing an energy-efficient computational paradigm in the future.