In this article, I will introduce a paper on machine learning using Born machines. A Born machine is a machine that learns using quantum mechanical existence probabilities rather than statistical probabilities. The method is roughly the same as Quantum Approximation Optimization Algorithm (QAOA) and Variational Quantum Eigensolver (VQE). We prepare a quantum circuit, put variables into clusters, and calculate the gradient from the covariance at the same time, aiming for minimum energy and zero covariance gradient. As a result of optimizing the Dots &amp; Bars dataset using this method, we found that the correct pattern can be obtained with a probability of about 92% with about 20,000 shots. This is a new method of quantum machine learning, but it did not look much like a quantum or classical machine learning method. In fact, except for the part where the covariance is calculated, it is not so different from the usual VQE method. I think I can implement only this part of the method, so I will try to implement it when I get a chance. <a href="https://arxiv.org/abs/1804.04168" rel="noopener noreferrer" target="_blank">[1804.04168] Differentiable Learning of Quantum Circuit Born Machine (arxiv.org)</a>

Paper Introduction: Born Machines.

Hikaru Wakaura