Abstract
We derive machine learning algorithms from discretized Euclidean fieldtheories, making inference and learning possible within dynamics described byquantum field theory. Specifically, we demonstrate that the $\phi^{4}$ scalarfield theory satisfies the Hammersley-Clifford theorem, therefore recasting itas a machine learning algorithm within the mathematically rigorous framework ofMarkov random fields. We illustrate the concepts by minimizing an asymmetricdistance between the probability distribution of the $\phi^{4}$ theory and thatof target distributions, by quantifying the overlap of statistical ensemblesbetween probability distributions and through reweighting to complex-valuedactions with longer-range interactions. Neural networks architectures areadditionally derived from the $\phi^{4}$ theory which can be viewed asgeneralizations of conventional neural networks and applications are presented.We conclude by discussing how the proposal opens up a new research avenue, thatof developing a mathematical and computational framework of machine learningwithin quantum field theory.