Merging multiple nodal networks
AssigneeD5Ai LLC
InventorsJames K. Baker, Bradley J. Baker
Life of this patent
Estimating the term…
Abstract
A system and method for controlling a nodal network. The method includes estimating an effect on the objective caused by the existence or non-existence of a direct connection between a pair of nodes and changing a structure of the nodal network based at least in part on the estimate of the effect. A nodal network includes a strict partially ordered set, a weighted directed acyclic graph, an artificial neural network, and/or a layered feed-forward neural network.
Claims39
A computer-implemented method for merging first and second nodal networks to create a merged network, wherein each of the first and second nodal networks comprises a plurality of layers, wherein each of the layers of the first and second nodal networks comprise one or more nodes, wherein at least some of the nodes of the first nodal network are connected by first network arcs, wherein at least some of the nodes of the second nodal network are connected by second network arcs, and wherein each of the first and second networks are, at least partially, trained such that each of the first and second network comprise knowledge, the method comprising:
- merging, by a computer system, the first and second nodal networks into the merged network such that, prior to training of the merged network, no node in the first nodal network covers a node in the second nodal network and no node in the second nodal network covers a node in the first nodal network;
- evaluating, by the computer system, potential cross-connections between the first and second nodal network in the merged network, wherein each potential cross-connection is an arc between a node in the first nodal network and a node in the second nodal network, and wherein the evaluation of the potential cross-connection is based on an estimated improvement in an objective of the merged network that includes the evaluated potential cross-connection;
- adding, by the computer system, at least one potential cross-connection to the merged network based, at least in part, on the evaluation such that, after the at least one potential cross-connection is added to the merged network, a node in the second nodal network covers a node in the first nodal network;
- and
- after adding the at least one potential cross-connection to the merged network, training, by the computer system, the merged network.
A computer system for merging first and second nodal networks to create a merged network, wherein each of the first and second nodal networks comprises a plurality of layers, wherein each of the layers of the first and second nodal networks comprise one or more nodes, wherein at least some of the nodes of the first nodal network are connected by first network arcs, wherein at least some of the nodes of the second nodal network are connected by second network arcs, and wherein each of the first and second networks are, at least partially, trained such that each of the first and second network comprise knowledge, the computer system comprising:
- one or more processor cores;
- and
- a memory in communication with the one or more processor cores, wherein the memory stores software that when executed by the one or more processors cause the one or more processor cores to:
- merge the first and second nodal networks into the merged network such that, prior to training of the merged network, no node in the first nodal network covers a node in the second nodal network and no node in the second nodal network covers a node in the first nodal network;
- evaluate potential cross-connections between the first and second nodal network in the merged network, wherein each potential cross-connection is an arc between a node in the first nodal network and a node in the second nodal network, and wherein the evaluation of the potential cross-connection is based on an estimated improvement in an objective of the merged network that includes the evaluated potential cross-connection;
- add at least one potential cross-connection to the merged network based, at least in part, on the evaluation such that, after the at least one potential cross-connection is added to the merged network, a node in the second nodal network covers a node in the first nodal network;
- and
- after adding the at least one potential cross-connection to the merged network, train the merged network.
Dependencies are inferred from the claim text.