Study of finite-size corrections in systems with a large number of degrees of freedom, by means of neural networks.