Abstract
This paper introduces the quantum deep sets model, expanding the quantummachine learning tool-box by enabling the possibility of learning variadicfunctions using quantum systems. A couple of variants are presented for thismodel. The first one focuses on mapping sets to quantum systems through statevector averaging: each element of the set is mapped to a quantum state, and thequantum state of the set is the average of the corresponding quantum states ofits elements. This approach allows the definition of a permutation-invariantvariadic model. The second variant is useful for ordered sets, i.e., sequences,and relies on optimal coherification of tristochastic tensors that implementproducts of mixed states: each element of the set is mapped to a densitymatrix, and the quantum state of the set is the product of the correspondingdensity matrices of its elements. Such variant can be relevant in tasks such asnatural language processing. The resulting quantum state in any of the variantsis then processed to realise a function that solves a machine learning tasksuch as classification, regression or density estimation. Through syntheticproblem examples, the efficacy and versatility of quantum deep sets andsequences (QDSs) is demonstrated.