Ponente
David Garófalo Aedo
(Facultad de Física, Universidad de La Habana)
Descripción
A Restricted Boltzmann Machine (RBM) is a generative neural network made of a layer of visible neurons fully
connected to a single layer of hidden neurons. RBMs are called “restricted” because there are not connections
between units within the same layer. This model is typically used to learn a statistical representation of a
dataset in order to generate new samples from the same distribution. In this work, we explore the relationship between the structure of the training dataset and the quality of the new samples generated by RBM. We train multiple networks on the MNIST dataset and present numerical results that suggest that RBM might have a preference towards some type of data when generating new samples.
Autores primarios
Cristina Díaz Faloh
(Facultad de Física de la Universidad de la Habana)
David Garófalo Aedo
(Facultad de Física, Universidad de La Habana)
