Optimization Algorithms for Quantum Deep Learning
PhD Candidate: Guillaume Verdon-Akzam
Abstract: Quantum variational algorithms, also known as quantum neural networks (QNN’s), have seen a recent surge of interest, namely, due to their partial robustness to noise, and due to the feasibility of their implementation on near-term quantum devices. Such variational algorithms generally involve the execution of parameterized quantum circuits along with the optimization of the parameters such as to minimize some cost functional of the output state of the circuit. This optimization of the parameters of the circuit can be considered as the training of the quantum neural network. Typically, this QNN training is done using classical black box optimization algorithms running on a classical computer which is used as a co-processor in conjunction with a quantum computer. In this talk, I will introduce novel quantum-enhanced optimization algorithms to perform such optimization tasks using quantum dynamics on both digital and analog quantum computers, as well as novel classical algorithms surpassing the performance of the current state-of-the-art classical methods for quantum neural network training.