Representational Power of Restricted Boltzmann Machines and Deep Belief Networks
Representational Power of Restricted Boltzmann Machines and Deep Belief Networks is a scholarly work by Yoshua Bengio, published in 2008 in ''Neural Computation''. The main subjects of the publication include theoretical computer science, block, computer science, power, machine learning, mathematics, probabilistic logic, restricted Boltzmann machine, Boltzmann machine, deep learning, inference, adversarial machine learning, generative model, generative grammar, transfer learning, generative adversarial network, artificial intelligence, artificial neural network, deep belief network, and Boltzmann constant. The authors first prove that adding hidden units yields strictly improved modeling power, while a second theorem shows that RBMs are universal approximators of discrete distributions.