Restricted Boltzmann Machines

Home Glossary Item Restricted Boltzmann Machines
« Back to Glossary Index

Restricted Boltzmann Machines (RBMs) are a type of artificial neural network that falls under the category of generative models. They are composed of visible and hidden units arranged in a bipartite graph structure, where connections exist only between visible and hidden units and not within the same group. RBMs are used for tasks such as dimensionality reduction, feature learning, and collaborative filtering, and they are particularly adept at unsupervised learning tasks.


RBMs are the ability to learn intricate patterns and relationships within data. RBMs utilize a stochastic approach, where the connections between units are assigned weights that are iteratively adjusted through a process called contrastive divergence during training. The learning process aims to minimize the energy difference between the model’s reconstructed data and the original input data. RBMs are considered generative models because they can generate new data samples that are similar to the data they were trained on, making them useful for tasks like data denoising and generation.


Despite their powerful capabilities, training RBMs can be computationally demanding, and they have been largely surpassed by more advanced models such as deep belief networks (DBNs) and variational autoencoders (VAEs). Nevertheless, the essence of RBMs contributes to the understanding of generative models and their role in uncovering hidden patterns in complex data.

« Back to Glossary Index