Beginning Anomaly Detection Using

restricted Boltzmann machines (RBM), deep Boltzmann machines 

(DBM), and deep belief networks (DBN) are much more suitable and practical to work 

with, although they are a bit outdated and have no support from the major frameworks 

such as Keras, TensorFlow, and PyTorch. Despite that, they still see some new uses today, 

even though they are overshadowed by newer deep learning models. For our purposes, we 

will look at applying the 

RBM to anomaly detection, particularly because it is the easiest of 

the three Boltzmann machine derivations to implement and because it is simpler to work 

with when we consider the mathematics (which are still at an advanced level) at play.

F

E

D

C

B

A

H

G

Hidden Nodes

Visible Nodes

W

BC  

is the weight 

between nodes B and 

C

W

GF

Figure 5-1.  A graph showing how a Boltzmann machine can be structured. Notice 

that all of the nodes are interconnected, even if they are in the same layer

Chapter 5   Boltzmann maChines

---

181