A Variational Algorithm for Quantum Single Layer Perceptron

Hybrid quantum-classical computation represents one of the most promising approaches to deliver novel machine learning models capable of overcoming the limitations imposed by the classical computing paradigm. In this work, we propose a novel variational algorithm for quantum Single Layer Perceptron (qSLP) which allows producing a quantum state equivalent to the output of a classical single-layer neural network. In particular, the proposed qSLP generates an exponentially large number of parametrized linear combinations in superposition that can be learnt using quantum-classical optimization. As a consequence, the number of hidden neurons scales exponentially with the number of qubits and, thanks to the universal approximation theorem, our algorithm opens to the possibility of approximating any function on quantum computers. Thus, the proposed approach produces a model with substantial descriptive power and widens the horizon of potential applications using near-term quantum computation, especially those related to quantum machine learning. Finally, we test the qSLP as a classification model against two different quantum models on two different real-world datasets usually adopted for benchmarking classical algorithms.