| Abstract: | I consider how techniques in machine learning and in quantum information processing can have direct impacts upon each other. From the classical machine learning perspective, I describe our efforts to automate key procedures in semiconducting quantum computing experiments by the creation of a large synthetic, labeled data set via modeling which enables the use of convolutional neural networks to replace human intuition in tuning up experimental apparati. From the quantum information processing perspective, I consider how access to algorithms that create synthetic data sets can enable polynomial, superpolynomial, or even exponential improvements in the size of data set necessary to achieve the same level of accuracy in the reinforcement learning paradigm, when equipped with a fully functioning quantum computer.
|