pydcop.algorithms.maxsum¶
MaxSum: Belief-propagation DCOP algorithm¶
Max-Sum [FRPJ08] is an incomplete inference-based DCOP algorithm.
This is a synchronous implementation of Max-Sum, where messages are sent and received in cycles. For an asynchronous implementation, see. A-Max-Sum
Algorithm Parameters¶
- damping
amount of dumping [0-1]
- damping_nodes
nodes that apply damping to messages: “vars”, “factors”, “both” or “none”
- stability
stability detection coefficient
- noise
noise level for variable
- start_messages
nodes that initiate messages : “leafs”, “leafs_vars”, “all”
FIXME: add support for stop_cycle
Example
pydcop solve -algo maxsum \
--algo_param stop_cycle:30 \
-d adhoc graph_coloring_csp.yaml
FIXME: add results
See also
A-Max-Sum: an asynchronous implementation of Max-Sum.