Multinode model is the concept of using an individual actor or a pair of actors in multiple contexts where producing adaptive actions is necessary. The contexts might relate to the components of a system you develop or entities external to the system. The nodes of a multinode model are adaptive programmatic representations of those contexts. They provide learning capabilities to the system, the ability to produce actions based on learned interactions, and can define preprogrammed behavior. A special case of multinode model is a single-node model.
A node possesses control when it performs interactions with other entities by executing the instructions of a possibly nondeterministic subroutine contained in the node. The execution of this subroutine we briefly call node execution. At present, QSMM supports not more than one node in a multinode model possessing control at any moment of time; in addition to that, a node can execute not more than one instruction at any time. A node can pass control to another node while executing an instruction at some location in the subroutine, and the other node can return control to that location afterwards.
A node learns something when the subroutine becomes more deterministic or stable. This added determinism affects subsequent interactions of this node with other entities.
Considering the sources of deterministic and stochastic behavior, their location, and their ability to influence each other gives answers to the questions “Which entity is choosing actions?” and “Is it easy for that entity to choose actions?” For example, a node can use a stochastic physical process, such as the interference of individual electrons, to produce random numbers for the subroutine to choose actions. If the node is not using a stochastic physical process, and hardware for executing the subroutine operates deterministically, then information the node receives from the environment may define the choice of actions.
Working with a multinode model begins with its preparation. It includes registering meta-classes for instructions executable by nodes and registering the sets of instruction classes—different nodes can have different instruction sets. The multinode model learns and performs useful work during execution phase—the execution of nodes and transferring control between them. Various information about a multinode model is available at various points in its lifetime.
|• Principle of Operation|
|• Creating a Multinode Model|
|• Executing a Multinode Model|
|• Listing a Multinode Model|
|• Error Handling for a Multinode Model|