Quantum Neural Network for Image Classification Using TensorFlow Quantum

Abstract

Quantum neural networks are artificial neural networks that are designed using the principles of classical artificial neural networks and quantum information. Developing a quantum neural network requires knowledge of both classical neural networks and quantum computing techniques. TensorFlow Quantum is a library for developing quantum and hybrid neural network models. The MNIST handwritten digits dataset was used as a sample dataset for quantum neural network development. This chapter introduces the design and development process of a simple quantum neural network for image classification tasks using the TensorFlow quantum library. Also, it compared the classification performances of the quantum neural network and classical neural network on the MNIST handwritten digits classification. This chapter identified that the quantum neural network performed better than the classical neural network on digit classification. Also, the chapter discussed the advantages and limitations of quantum neural networks in image classification. The first section of the chapter introduced the quantum neural network models and the TensorFlow quantum library. Afterward, the data preparation steps such as data loading, downscaling, contradictory removal, and Tensorflow Quantum circuit conversions were discussed. Furthermore, the building process of quantum neural networks for image classification, such as quantum neural network designing, the model circuit to the Tensorflow Quantum model binding, and model training, was discussed in the third section of the chapter. Subsequently, the testing performance of the simple quantum neural network was discussed in section four of this chapter. Finally, the conclusions of the chapter were discussed in the fifth section.