Initially visible models of subnetwork construction algorithms in AGTIntellegence networking

©P.Drobyshev, D.Novhorodkina

Visible initial models of subnetwork construction algorithms in AGTIntellegence network.

Let’s call it:

AGTInw (Application) – a gadget application program that connects gadgets and people working on them via the internet, hereinafter referred to as an application.

AGTIntelligence network (hereinafter – the Network) – a subnetwork on the Internet that connects people and gadgets with the AGTInw application installed in order to receive or provide work on the Network.

AGTIneuron (Agent) – association in the Network operation of a person (hereinafter referred to as the Owner, AGTIowner) operating on the Network on the AGTInw application with its Gadget.

AGTIgadget (AGTIgadget, Gadget) – a gadget connected to the Network and operated by the Owner

SubnetAGTI – An association of Agents that perform a specific task.

Simple program (simple procedure) – a parametric program which, with different parameters, can be loaded simultaneously many times with different parameters (the number of times required in the Subnetwork) on one Gadget and all this number is successfully and quickly executed by the Gadget.

Goalset – an expression of will of the Agent(s) to create or participate in the work of the Network

AGTIsynapse – A connection between two Agents that has its own weight, and possibly its own simple program of operation. The Agent accepts (installs in the Application on his/her Gadget) the Synapse either automatically or manually (at the Owner’s will).

AGTIsynmatrix – If three or more Agents are connected by a simple program P(A,B,C,…) similar to a Synapse, but connecting several Agents at once (term ©V.Bubnov).

The weight (AGTIpower) of an Agent is a tensor function (of time and other parameters) available for each Agent, different at each moment, available to all Agents (possibly partially), and serving for clustering of the Network by targeting.

Agent clustering

In order for the Agents creating the Subnetwork to be able to recruit the necessary Agents to each layer of their Subnetwork, it is reasonable to create a methodology for labeling Agents with Weights – tensor functions (of time and other parameters). Agent A corresponds to T(A).

It’s like this:

– The Agent owner can work in AGTInw now and for two more hours

– The owner of the Agent has lived in Poland for more than two years

– The Agent gadget is ready to provide 1Gb of RAM to work in AGTInw

– Payment for 100 hours of Gadget operation not more than 5 AGTI

And other…

The degrees of access different Agents have to Weight may vary.

Building a traditional architecture subnetwork

A group of Agents (usually connected by a synapse-matrix) creating a Subnetwork P selects the target parameters of P, i.e., its practical functions, a participation payment C(T(A)) and a performance bonus PR(A), the participants of the layers are selected (by their acceptance) according to the parameter weights appropriate for each layer, thus clustering the Subnetwork.

Within and between the layers of the Subnetwork are created Synapses P(A,B) – tensors in the form of simple programs, possibly generated automatically by algorithm by the creators of the Subnetwork for each Synapse, P(A,B) are loaded into Gadgets A and B, and may be different, but guarantee fast interaction between A and B. The interaction in a given Subnetwork between A and B goes through a particular Synapse P(A,B).

Thus, a neural network of traditional architecture is easily specified by the functions of synapses in the layers clustered by Agent Weights.

Unconventional architecture.

P(A,B) may have some stochastic parameter p with some probability distribution of values (at a given time, or from an address in the Network, Subnetwork, etc.). Then the Subnet itself begins to behave probabilistically, like a living brain. Let us call such a subnetwork stochastic.

The architecture of stochastic networks with synapse matrices is subject to study, AGTIntellegece network provides a full opportunity to experiment, study and utilize such Sub-networks.

Fractal network. Agent A1, a member of subnetwork P1 to realize the goal setting of P1 creates Subnetwork P2, further. In it, agent A2 creates P2 and so several layers.

Thus, the methodology of building Subnetworks in AGTIntelligence network is described in general and in further articles we will specify already approximate algorithms that will launch the Network into useful work for people.