CAUSAL DISCOVERY AND MISSING VALUE IMPUTATION

A computer-implemented method comprising: receiving an input vector comprising values of variables; using a first neural network to encode the variables of the input vector into a plurality of latent vectors; inputting the plurality of latent vectors into a second neural network comprising a graph neural network, wherein the graph neural network is parametrized by a graph comprising edge probabilities indicating causal relationships between the variables, in order to determine a computed vector value; and tuning the edge probabilities of the graph, one or more parameters of the first neural network and one or more parameters of the second neural network to minimise a loss function, wherein the loss function comprises a measure of difference between the input vector and the computed vector value and a function of the graph.