A Comprehensive Review of Data Encoding Techniques for Quantum Machine Learning Problems
Mandaar B. Pande
Abstract
In this paper, we focus on providing insights into various data encoding techniques for solving real-life problems using quantum machine learning, specifically for gate based quantum computers. At present, quantum computers have small qubits numbers, are noisy and are termed as NISQ (Noisy Intermediate Scale Quantum) computers. Generally, there are four steps involved in solving quantum machine learning problems. First is to encode data into a quantum computer using various data encoding techniques. The second step is to create a model of the system being addressed which is essentially the development of a quantum circuit using 1- and 2- qubit quantum gates. The third step is to train the model using gradient based or other optimization techniques. The fourth step is measurement and analysis of classical outputs of the model and improve various metrics of the model to make it robust for providing accurate results. In this research work, we evaluate and provide the mathematical underpinnings of different data encoding techniques which are used in today's NISQ quantum computers. Further, we discuss performance and efficiency of the various data encoding techniques.