Quantum Computing-Enabled Machine Learning for an Enhanced Model Training Approach
Book Chapter
Soni, J, Prabakar, N, Upadhyay, H. (2023). Quantum Computing-Enabled Machine Learning for an Enhanced Model Training Approach
. 1085 201-216. 10.1007/978-981-19-9530-9_12
Soni, J, Prabakar, N, Upadhyay, H. (2023). Quantum Computing-Enabled Machine Learning for an Enhanced Model Training Approach
. 1085 201-216. 10.1007/978-981-19-9530-9_12
Machine learning is an exciting area where ever-growing problems such as anomaly detection using sensor data, natural language processing, image processing, etc., are solved using complex yet fascinating algorithms. Such algorithms learn the function that maps input to output from the training examples. The algorithms are evaluated with the validation data and then used to predict the output for an unknown dataset. For the past few years, scholars have been researching to improve such classical machine learning algorithms using quantum computing. Some of the latest research is the optimization of computationally expensive algorithms with quantum computing and transforming stochastic procedures into the semantics of quantum theory. The quest for the learning-based algorithm is aspiring: the discipline seeks to comprehend what learning is and studies how algorithms approximate learning. Quantum machine learning takes these aspirations further by looking at the subatomic level to aid learning. Machine learning-based algorithms minimize a constrained multivariate function. Different algorithms have different hyper-parameters that need to be tuned for the trained model to generalize well. Such optimization has high time and space complexity, which is central to learning theory. This contribution gives an organized overview of the evolving arena of quantum machine learning. It presents the methods as well as practical details. We start this chapter by introducing the major components of Quantum Computers, where we provide an overview of quantum computing with an in-depth explanation of the superposition of state, which will be crucial for all quantum algorithms. Next, we exploit a fascinating phenomenon called entanglement in quantum computations. Parallelism is the key to speeding up the training process of learning algorithms. One of the significant advantages of quantum computing is quantum parallelism which we will explore through Grover's search algorithm. Next, we go over the learning mechanism of a traditional machine learning algorithm, namely a Support Vector Machine (SVM) trained on classical computers. Furthermore, we exploit the Quantum SVM (QSVM) trained on quantum computers. Finally, we solve the famous malignant breast cancer classification problem using the quantum SVM algorithm. We study various simulations in Python language on the mentioned dataset to analyze their time complexity and performances on standard evaluation metrics, namely accuracy, precision, recall, and F1-score. Two simulations are conducted using classical machine learning algorithms using the Python library Scikit-learn. Finally, the last simulation is based on IBM's real quantum computer using its quantum machine learning library called Qiskit.