(This article does not explain the basic concepts of the theory of neural networks.) For those who are not familiar with them, I advise you to read before reading to avoid further misconceptions.)
 
 
The essence of this text is acquaintance with some types of neural networks, which are covered in the Russian-speaking spaces, it is not so often, if not to say that at all, extremely rare.
 
www.lund.irf.se/HeliosHome/rulebased.html
Now the question is how to make a learning neural network out of this rigorous system?
 
 
First, an important point is to add weight to each structure, to each edge. Each weight will reflect the probability of the ratio of one or another element to the group of others (for example, the input parameter A to the first neuron of the hidden layer, or to the group AB, respectively), or to the answer X, Y, Z, etc.
 
 
Perhaps the reader will not be entirely clear where such neural networks can be useful, in which case I will give a fairly simple example:
 
 
Suppose we do not have a large sample of data, but only "
? a generalized [/i] opinion." We want to create a neural network that would give out an individual menu for a person.
 
 
Suppose that we do not know anything about the tastes and preferences of this user, but we still need to begin with something. We are generalized typical menu scheme:
 
 
 
breakfast-omelette
 
lunch-soup
 
dinner porridge
 
 
Accordingly, in the early days a person receives just such a menu, but with an "acquaintance" of the neural network with the user's preferences, the weight binding breakfast and omelet becomes smaller, and the weight binding breakfast and porridge increases. Accordingly, now, the neural network is "clear", what exactly the user prefers in this or that meal (in this case, it turns out that our user likes porridge for breakfast rather than omelet). Over time, perhaps a person's preferences will change and the neural network will again adjust to them.
 
 
So. At a minimum, RBNN networks can be very useful in cases where there are no large samples, when there are none at all, and also when we need a system that fully adapts to a particular person. Moreover, such neural networks are quite simple, which allows them to be used for training other people and an understandable explanation of the actions of neural networks.
 
 
Earlier it was always accepted to say that the neural network is a "black box", and everything that is inside it can not be explained in an accessible way. Now, having the above structure, it is possible to build a neural network that would not only be effective, but also accessible to the understanding of its surrounding mechanisms.