A multi-layer network planning system can determine a set of regenerator sites (RSs) that have been found to cover all paths among a set of nodes of an optical layer of a multi-layer network and can determine a set of candidate RSs in the optical layer for use by the links between a set of nodes of an upper layer, wherein each RS can be selected as a candidate RS for the links. The system can determine a binary path matrix for the links between the set of nodes of the upper layer. The system can determine a min-cost matrix that includes a plurality of min-cost paths. The system can determine a best RS from the set of candidate RSs and can move the best RS from the set of candidate RSs into the set of RSs for the links. The system can then update the binary path matrix.

A network management system can be configured to identify routes for satisfying a set of demands on a communications network using layer graph(s). The network management system can generate the layer graph(s) using a network graph that represents the optical communication network and an associated sets of available frequency slots. The network management system can iteratively identify candidate path(s) on the layer graph(s) that correspond to each of the demands and determine a cost for each candidate path. The cost for a candidate path can depend on a set of available edges affected by the selection of the candidate path. In each iteration, the network management system can select the lowest cost candidate path, update the network graph to reflect the selection of this candidate path, update the layer graph(s) based on the updating of the network graph, and update the candidate paths for the remaining demands as needed.

A method for optimizing network performance using reinforcement learning (RL) agents is disclosed. The method includes identifying multiple network segments within a network, each including network nodes; generating and training respective RL agents for at least a subset of these segments based on performance metrics indicative of data flow within each segment, independently of specific segment topology information; receiving outputs from the trained RL agents, including policies or performance evaluations; generating recommendations based on the received outputs; and causing network actions to be implemented based on these recommendations. In various embodiments, the RL agents utilize metrics such as Quality of Service (QOS), Quality of Experience (QoE), or radio resource management parameters. Recommended actions may include switching traffic paths, adjusting wireless parameters, and proactively preventing network congestion to enhance network operation and user experience.

Various embodiments provide for clock signal distribution within a processor, such as a machine learning (ML) processor, using a photonic fabric.