The Restricted Boltzmann Machine (RBM) is an unsupervised learning artificial neural network. It is a generative model that can learn a probability distribution from a set of input data.

"There is no reason and no way that a human mind can keep up with an artificial intelligence machine by 2035." - Gray Scott.

Although the Boltzmann machine is named after Austrian scientist Ludwig Boltzmann, who established the Boltzmann distribution in the twentieth century, Stanford scientist Geoff Hinton invented this form of the network. Hinton and Salakhutdinov created RBM in the mid-2000s as a solution to the problem of unsupervised learning. It has two layers of neurons—one visible and one buried. The hidden layer represents the network's learned characteristics, while the visible layer represents input data.

What is a Restricted Boltzmann Machine?

The RBM is referred to as "restricted" because connections between neurons in the same layer are not permitted. In other words, each visible layer neuron is only linked to neurons in the hidden layer, and vice versa. By lowering the dimensionality of the input, the RBM can learn a compressed version of the input data. 

In contrast, "unrestricted" Boltzmann machines may contain concealed unit connections. This restriction enables training algorithms that are more efficient than those available for the general class of Boltzmann machines, specifically the gradient-based contrastive divergence algorithm.

Types of Boltzmann machines

• Restricted Boltzmann Machines (RBMs)
• Deep Belief Networks (DBNs)
• Deep Boltzmann Machines (DBMs)

Application

The Restricted Boltzmann Machine approach, used for feature selection and extraction, is critical in the age of Machine Learning and Deep Learning for dimensionality reduction, classification, regression, and a variety of other applications.

RBMs have applications in 

• dimensionality reduction, 
• classification, 
• collaborative filtering, 
• feature learning, 
• topic modelling, and 
• body quantum mechanics. 

Depending on the assignment, they can be instructed under supervision or unsupervised.

Why Boltzmann machines?

Boltzmann machines are commonly used to solve various computing issues, such as search problems, where the weights on the connections can be specified and utilized to represent the cost function of the optimization problem. The Boltzmann Constant is used in classical statistical mechanics to express the equipartition of an atom's energy. It means the Boltzmann factor. It's essential for statistical entropy. Semiconductor physics uses thermal voltage.

In Machine Learning and Deep Learning, the Restricted Boltzmann Machine technique for feature selection and extraction is essential for dimensionality reduction, classification, regression, and other applications.

Conclusion

A Boltzmann machine is a stochastic spin-glass model with an external field, a stochastic Ising model. A restricted Boltzmann machine has no intralayer connection. Therefore nodes in the same layer are independent.

Furthermore, RBM is an undirected, energy-based graphical model. It is widely utilized in both unsupervised and supervised machine learning. RBM is often trained via contrastive divergence (CD).

Image source: Unsplash