# restricted Boltzmann machine

> Boltzmann machine whose neurons form a bipartite graph with visible and hidden neurons

**Wikidata**: [Q7316287](https://www.wikidata.org/wiki/Q7316287)  
**Wikipedia**: [English](https://en.wikipedia.org/wiki/Restricted_Boltzmann_machine)  
**Source**: https://4ort.xyz/entity/restricted-boltzmann-machine

## Summary
A restricted Boltzmann machine (RBM) is a stochastic neural network and a specific type of Boltzmann machine. It is distinguished by its architecture, in which neurons form a bipartite graph divided into visible and hidden units. Invented in 1986 by Paul Smolensky and Geoffrey Hinton, it is also known as a "Harmonium."

## Key Facts
- **Inventors:** Paul Smolensky and Geoffrey Hinton.
- **Invention Date:** 1986.
- **Architecture:** Uses a complete bipartite graph structure separating visible and hidden neurons.
- **Classification:** A subclass of both the Boltzmann machine and the stochastic neural network.
- **Acronym:** RBM.
- **Alternate Name:** Harmonium.
- **Visual Schematics:** Available in English, French, Ukrainian, and Russian (via Wikimedia Commons).
- **Wikipedia Presence:** Available in 10+ languages including English, Spanish, French, Russian, and Korean.

## FAQs
### Q: Who invented the restricted Boltzmann machine?
A: The restricted Boltzmann machine was invented in 1986 by the cognitive scientist Paul Smolensky and the computer scientist Geoffrey Hinton.

### Q: How does a restricted Boltzmann machine differ from a standard Boltzmann machine?
A: An RBM is distinguished by its specific structure, which forms a bipartite graph of visible and hidden neurons (often implemented as a complete bipartite graph), whereas standard Boltzmann machines may have more complex connectivity.

### Q: What is the scientific classification of an RBM?
A: It is classified as a type of stochastic neural network and a subclass of the Boltzmann machine.

## Why It Matters
The restricted Boltzmann machine represents a critical architectural simplification in the field of artificial neural networks. By restricting the connectivity of a standard Boltzmann machine into a bipartite graph of visible and hidden neurons, the model provides a structured approach to representing probability distributions.

This architecture is significant because it organizes neurons into distinct visible and hidden layers, allowing for the modeling of complex data structures through stochastic processes. The involvement of Geoffrey Hinton, a pioneer in the field, underscores its historical importance in the development of artificial intelligence research. Its presence across multiple global encyclopedias and technical stacks (such as Stack Exchange) highlights its enduring relevance as a foundational concept in machine learning.

## Notable For
- **Bipartite Architecture:** Its defining characteristic is the use of a complete bipartite graph that strictly separates visible and hidden neurons.
- **Historical Pedigree:** Co-invented by Geoffrey Hinton, a major figure in AI research, and Paul Smolensky in 1986.
- **Terminology:** It is uniquely referred to as a "Harmonium" in certain academic contexts.
- **Global Standardization:** It is indexed in the Encyclopedia of China (Third Edition) and maintains active tags on technical platforms like Stack Overflow and AI Stack Exchange.

## Body

### Definition and Structure
The restricted Boltzmann machine (RBM) is a stochastic neural network. It is defined structurally as a Boltzmann machine whose neurons form a bipartite graph. This architecture divides the units into two distinct groups:
*   **Visible Neurons**
*   **Hidden Neurons**

The structure utilizes a **complete bipartite graph**, meaning every neuron in the visible layer is connected to every neuron in the hidden layer, though the source text does not specify the nature of information flow (e.g., undirected) beyond the graph topology.

### Invention and History
The RBM was invented in **1986**. It is credited to two primary researchers:
*   **Paul Smolensky**
*   **Geoffrey Hinton** (described in sources as a British-Canadian computer scientist, psychologist, and AI researcher).

### Classification and Identifiers
The entity is formally classified as a subclass of the **Boltzmann machine** and the broader category of **stochastic neural networks**.

**Key Identifiers and Aliases:**
*   **Short Name:** RBM
*   **Alternate Name:** Harmonium
*   **Wikidata ID:** Q175051 (implied by structured property context)
*   **Freebase ID:** /m/0ndhjdc
*   **Microsoft Academic ID:** 199354608 (discontinued)
*   **Encyclopedia of China (Third Edition) IDs:** 112972, 509893

### Resources
The entity is widely documented across multiple languages and platforms:
*   **Wikipedia:** Available in languages including English, French, Spanish, Italian, Russian, Ukrainian, Korean, Vietnamese, Farsi, and Catalan.
*   **Technical Communities:** Active tags exist on Stack Overflow (`rbm`) and AI Stack Exchange (`restricted-boltzmann-machine`).
*   **Visuals:** SVG schematics are available for the model in English, French, Ukrainian, and Russian.

## References

1. [OpenAlex](https://docs.openalex.org/download-snapshot/snapshot-data-format)