My research project consists in modeling the link between beauty and neural fluency. For this, different objectives have been established. First thing I had to do is to add a sparsity constraint in the architecture of a convolutional neural network (CNN) model previously trained to make it biologically plausible. In a second step, I specialize the model on a specific task which is the recognition of faces of an ethnic group compared to a beauty score. Based on a concept in psychology which is the perceptual fluidity. This theory states that individuals show a generalized preference for familiar stimuli. If the trained model is specialized in recognizing individuals, will this increase the model's ability to predict the beauty of individuals?