Open quantum systems composed of atoms interacting with light exhibit behaviour that is akin to that of associative memories [1]. This means that they possess stationary states that can be interpreted as stored memory patterns, which are retrieved when the initial state is inside the basin of attraction of a given pattern [2]. The corresponding pattern retrieval dynamics can be observed in actual experimental settings. In these experiments atoms are confined within an optical cavity whose photons mediate long-range interactions [3]. Stored patterns are encoded in the atom-light coupling constants. This setting offers an interesting opportunity for studying quantum generalisations of associative memories and stored (quantum) patterns in this context [4]. Moreover, it allows to systematically construct scenarios in which quantum effects might be beneficial, e.g., for speeding up the pattern retrieval process [5]. I will talk about recent research of my group on this subject, which builds a bridge between classic machine learning concepts, such as the Hopfield Neural Network, and most recent experimental manifestations of synthetic quantum matter.

[1] E. Fiorelli et al., Physical Review Letters 125, 070604 (2020)

[2] F. Carollo and I. Lesanovsky, arXiv:2009.13932 (2020)

[3] V.D. Vaidya et al., Physical Review X 8, 011002 (2018)

[4] P. Rotondo et al., Journal of Physics A 51, 115301 (2018)

[5] E. Fiorelli et al., Physical Review A 99, 032126 (2019)