Quantum error correction using quantum convolutional neural network

Abstract

Quantum computing with its powerful attributes - entanglement and superposition, is revolutionizing modern computation. Moreover, quantum computation can be applied to a wide range of real-world applications, including cybersecurity and cryptography, computational chemistry, artificial intelligence, prime factorization, and so on. However, the biggest impediment of quantum computation is quantum error. Often the decoherence caused by the environment or any other malfunction creates quantum errors which can arbitrarily change the state of a quantum system and destroying its information content - bit flip & phase flip error, unwanted measurement error, etc. A specific Quantum Error Correction (QEC) code can rectify some particular errors. But, it is usually not optimized for any random error. As of today, ‘Deep Learning’ techniques in analyzing data have been very promising which is influencing researchers to apply these methods to ‘Quantum Computation’ problems. We have implemented a Quantum Convolutional Neural Network (QCNN) using Parameterized Quantum Circuit (PQC) on IBM’s open source Quantum Computing framework QISKIT. The generic structure of the QCNN consists of variational forms of the encoder and decoder of the error correction code, which is optimized during training. In this way, it constructs a quantum error correction method for a certain error model. We were able to retrieve quantum states with as much as 90% fidelity rate from the experiments. As a result, our model achieves a high fidelity while using relatively few parameters, which can be generalized for any error model.