I understand that our models should provide as an output a vector of two dimensions, but how should we consider it? [1,0], for example, should correspond to "healthy" or "unhealthy"