pydcop.distribution.oneagent

The oneagent distribution algorithm assigns exactly one computation to each agent in the system.

It is the most simple distribution and, when used with many DCOP algorithms, it replicates the traditional hypothesis used in the DCOP literature where each agent is responsible for exactly one variable.

Note that this applies to algorithms using a computation-hyper graph model, like DSA, MGM, etc.

The oneagent distribution does not define any notion of distribution cost.

Functions

distribute(computation_graph: pydcop.computations_graph.objects.ComputationGraph, agentsdef: Iterable[pydcop.dcop.objects.AgentDef], hints: pydcop.distribution.objects.DistributionHints = None, computation_memory=None, communication_load=None, timeout=None) → pydcop.distribution.objects.Distribution

Simplistic distribution method: each computation is hosted on agent agent and each agent host a single computation. Agent capacity is not considered.

Raises an ImpossibleDistributionException

Parameters

  • computation_graph (a ComputationGraph) – the computation graph containing the computation that must be distributed

  • agentsdef (iterable of AgentDef objects) – The definition of the agents the computation will be assigned to. There must be at least as many agents as computations.

  • hints – Not used by the oneagent distribution method.

  • computation_memory – Not used by the oneagent distribution method.

  • computation_memory – Not used by the oneagent distribution method.

Returns

distribution – A Distribution object containing the mapping form agents to computations.

Return type

Distribution

distribution_cost(distribution: pydcop.distribution.objects.Distribution, computation_graph: pydcop.computations_graph.objects.ComputationGraph, agentsdef: Iterable[pydcop.dcop.objects.AgentDef], computation_memory: Callable[pydcop.computations_graph.objects.ComputationNode, float], communication_load: Callable[[pydcop.computations_graph.objects.ComputationNode, str], float]) → float

As the oneagent distribution does not define any notion of distribution cost, this function always returns 0.

Parameters

  • distribution –

  • computation_graph –

  • agentsdef –

  • computation_memory –

  • communication_load –

Returns

0

Return type

distribution cost