NTBE

Network Behavior Evolution

Gativus consider a few distinct phases in evolution of life:

1. Unicellular
2. Multicellular with Central nervous system, and
3. Social with verbal or symbolic information exchange.

Unicellular

Unicellular organism has all features of the organism in one cell. In Gativus terminology, all components are built in the single NDDI and internally interconnected (hard wired). All structure of the organism is coded in the cell's DNA and any change (even smallest one) has to be done through an evolution in form of micro mutations and follow up selection of useful one. The purpose (or target) of existence of such organisms, also coded in its DNA and is concentrated around simple survival of the individual.

The unicellular shows capability to learn, which can be attributed to machine learning, similar to technical neurons (like perceptrons). The survival and persistent existence of unicellular species today - is a fact of proof of learning efficiency, which allows organism to survive.

A purpose/target of the organism is inside the existence of the cell itself. There is no network behavior on this level, as soon as, all activity is inside the organism.

Multicellular

Substantial technical breakthrough happened when evolution switched to the synaptic level in the CNS of multicellular. Then, instead of creating the change to the DNA, the CNS can grow axon-synapse-dendrite chain to achieve new functionality. The total code of the animal got split between inherited DNA and acquired skills, apparently in synaptic connections. As more advanced species developed - bigger part of the code become acquired vs. genetically inherited. This feature gave multicellular ability for fast adaptation to environment change.

The purpose/target of existence probably also obtained synaptic additions in form of parameters and can alter the inherited genetic descriptor, sometimes quite substantially.

In technical Gativus vision, this level corresponds to informational relations (VARE) which could be established between value components of two NDDIs. The mechanism of creation of synaptic connection in biological organism remains unclear, but some features would be stipulated: (1) NDDI - has to be named to provide identified points of relations (in Gativus - UNON and LOCN names); (2) there should be an authority able to order to grow axon/dendrite and form the synapse (in Gativus - ALSA).

This level starts to show a common behavioral activity, based on hard wired synaptic connections.

Network

Next step in evolution, have happened when multicellular organism obtained the communication channel and consequently become able to relay some signals. Initially, those signal were extremely simple - like plain alarm, but later they become more and more complex. Simultaneously, a environment representation, instead of being genetically coded, become an dynamic object model build on synaptic layer* (Gativus hypothesis). With achievement of a language as communication means, organism become capable to relay complete entities of the object model* (Gativus hypothesis). This led ability to harmonize the object model which would be considered as a mutual object model. And consequently, the group can operate on the mutual model , multiplying the ability of the single member by the factor of a number of species.

Generally, the group's activity would be seen as autonomous activity of individuals at the beginning. Every organism has its own benefits in operating together, and has capability to determine its status in the group and adjust its behavior to the interest of the group. But they remains to be a number of autonomous individuals, just operating together for higher efficiency.

However, with a development of the evolution, group starts to show some features of individual organism, such as business identification (what is similar to individual member identification), its own legal statute, decision making organs, ets. Upon achieving the level of writing, (what is based on symbolic encoding of object model entities) those groups become capable to create its own v-components in form of books or documents. The books would be printed and distributed among members, forming VARE relations with one-to-many connections. Such v-component would have a lot of limitations, such as delay after the value update (printing and distribution cycle), etc.

Perceiving the group via reading a book (Chinese soldier with comrade Mao guide)

Ability to form virtual nodes with physical location of components in participating members - is current stage of evolution, when human group - represented as Gativus Alliance GAAL can form virtual Alliance instance ANOD which places its components in the member's physical network* (Gativus hypothesis).

The second achievement of the evolution - that cortical construction of neurons that can host not only the values of the instances, but also the behavior. Providing that cortical object model has this feature - biological neural network is capable to build dynamic (active) object model, where its behaviors can be seem as network one vs. autonomous behavior of the single instance/neuron. As soon as object model become dynamic - changes in this model should be developed immediately, and any change - leads to production of next change. Therefore, new type of connections between nodes in the network has to be accepted.

The reason for operation - the Target of the organism - is still can be seen as survival and propagation in territory, but as soon as it upgraded to the dynamic model (vs. DNA coded or synaptic model) - the Target become more flexible and can be changed fast and numerous times, depending on relations with other nodes of the network. In case, the Target of GAAL group is considered - it will be based on virtual relations and virtual components. But reasonable to assume, that Alliance instance ANOD has similar to physical group capability to amend its Target, and moreover, to amend the Targets of participating nodes.

Consequently, ANOD which is a node with exclusive rights for the network - has to host alliance ultimate Target (such as survival), and as soon as, this target consist of smaller intermediate targets - also host all of them. A brain, which is biological hosting device, has to provide technical capability to have such targets, as well as to change them and to assign targets to other nodes of alliance.

Meta-models

Current biological organisms (of higher mammals) are based on simultaneous use of all three meta-models: genetic code (DNA), synaptic and network.

Genetic code GCMM can be associated with Gativus Structure and provided as s-Relations.

Synaptic metamodel SYMM - is a collection of Gativus synapse analog - Value Relation VARE. The collection of VAREs forms the synaptic part of the common model.

Network meta-model NEMM is based on developing of Gativus targets GATA and achieving them.

Realisaiton of network behavior

Generally, it would be numerous solution how to create node's capability to support network behavior. However, Gativus NEMM architecture is based on:

• Gativus nodes NDDI structure and their relations;
• Assumption that any activity has to be made for a benefits or the organism;

To achieve the network target - NDDI has have an ability to define this target. The definition would be a target ontology and will be collected by pieces from other nodes. Target definitions GATA- are small targets itself which would be inherited or aggregated from parents. The code layer GCMM provides basic node's performance to allow them to enter the Relations (NERA) and develop synaptic layer SYMM, which is capable to host components required to define GATA.

Noteworthy, that all activity of the organism beyond synaptic layer - generally is connected with a move - to affect the environment and obtain its reaction. Therefore - it is behavior based, while SYMM play its role in determining subject and object of the behavior. The predicate can be defined by GCMM's s-relations, but a triple will have ontology based meaning, since its subject and object are products of SYMM

Due to complexity of GATA - it would not be unlimited amount of them inside the alliance. Than, all GATA can be interconnected with one-to_many connection. What means that all GATA inputs are connected to all other outputs of the alliance. (Alliance would not be a whole Gativus network - it would be a small group of nodes). Providing the learning of the digital weights for the inputs 0/1 - will provide the behavioral learning of the alliance. Every meaningful sets of weights in the alliance - will be registered as TCOM - Target common maps, where maps - are the weights.

Every alliance would have several TCOMs. As soon as it should be special appointed node for learning activity - ANOD - all TCOMs will be in in it.

Another feature of ANOD - is to develop the alliance network by creating the defined replica of TCOM which will represent future or past activity in achieving the ANOD ultimate target - survival and propagation.