Abstract: A method for modelling a shape of data in a GAN comprising a generator agent (G) for generating synthetic data (110, 220) and a discriminator agent (D) for distinguishing between the generated synthetic data (120, 220) and real original data (110) that follow an arbitrary, continuous or discrete, distribution defined by a n-dimensional vector of input variables xi. The method, before generating output synthetic data (220), computes an Inverse Smirnov transformation fSi?1 for each of the n input variables xi and attaches an activation function (200) to the generator agent (G), wherein the activation function (200) is a n-dimensional vector faG formed by the computed Inverse Smirnov transformations, faG=(fS1?1, fS2?1, . . . , fSn?1). The GAN implementing the method generates the output synthetic data (220) using the activation function (200) which outputs synthetic data (220) with a distribution (420, 420?) whose shape is the same as the arbitraty distribution of the original data (110).