Learning temporal data with variational quantum recurrent neural network

TL;DR

This paper introduces a variational quantum recurrent neural network model that leverages quantum circuits to learn and predict temporal data, demonstrating its effectiveness through numerical simulations on wave and spin dynamics.

Contribution

It presents a novel quantum recurrent neural network architecture using parametrized quantum circuits for temporal data learning, with analysis of interaction effects on performance.

Findings

Successfully predicts cosine and triangular waves

Shows quantum circuit's nonlinearity benefits temporal learning

Identifies optimal interaction strength range for performance

Abstract

We propose a method for learning temporal data using a parametrized quantum circuit. We use the circuit that has a similar structure as the recurrent neural network which is one of the standard approaches employed for this type of machine learning task. Some of the qubits in the circuit are utilized for memorizing past data, while others are measured and initialized at each time step for obtaining predictions and encoding a new input datum. The proposed approach utilizes the tensor product structure to get nonlinearity with respect to the inputs. Fully controllable, ensemble quantum systems such as an NMR quantum computer is a suitable choice of an experimental platform for this proposal. We demonstrate its capability with Simple numerical simulations, in which we test the proposed method for the task of predicting cosine and triangular waves and quantum spin dynamics. Finally, we analyze the dependency of its performance on the interaction strength among the qubits in numerical simulation and find that there is an appropriate range of the strength. This work provides a way to exploit complex quantum dynamics for learning temporal data.