Exponentially improved detection and correction of errors in experimental systems using neural networks

Abstract

We introduce the use of two machine learning algorithms to create an empirical model of an experimental apparatus, which is able to reduce the number of measurements necessary for generic optimisation tasks exponentially as compared to unbiased systematic optimisation. Principal Component Analysis (PCA) can be used to reduce the degrees of freedom in cases for which a rudimentary model describing the data exists. We further demonstrate the use of an Artificial Neural Network (ANN) for tasks where a model is not known. This makes the presented method applicable to a broad range of different optimisation tasks covering multiple fields of experimental physics. We demonstrate both algorithms at the example of detecting and compensating stray electric fields in an ion trap and achieve a successful compensation with an exponentially reduced amount of data.

Time-dependent atomic magnetometry with a recurrent neural network

“…NNs and their combinations in the so-called deep neural network (DNN) [17] have been used in quantum physics to model quantum dynamics [18][19][20], phase and parameter estimation [21][22][23][24], and quantum tomography [25,26], and they are increasingly employed in calibration and feedback tasks in quantum experiments [27][28][29][30]. To our knowledge, the present study is the first to use a deep neural network for the estimation of a stochastically varying perturbation such as a magnetic field interacting with a continuously monitored atomic system.…”