US 12,106,082 B2
Generative experiments for application deployment in 5G networks
Mudhakar Srivatsa, White Plains, NY (US); Dinesh C. Verma, New Castle, NY (US); Satish Sadagopan, Leawood, KS (US); Mathews Thomas, Flower Mound, TX (US); and Utpal Mangla, Toronto (CA)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed on May 20, 2021, as Appl. No. 17/326,278.
Prior Publication US 2022/0374215 A1, Nov. 24, 2022
Int. Cl. G06F 8/60 (2018.01); G06N 3/08 (2023.01); H04W 84/04 (2009.01)
CPC G06F 8/60 (2013.01) [G06N 3/08 (2013.01); H04W 84/042 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A computer implemented method for determining a placement of an application being added to a network, comprising:
determining a logical-physical mapping for the application being added to the network;
identifying one or more sub-graphs in the logical-physical mapping of the application being added to the network;
iteratively identifying a sub-graph with a lowest matching score between embedding vectors of a logical node la and a physical node pn from the one or more sub-graphs, wherein after each iteration, a next lowest matching sub-graph of the one or more sub-graphs is selected;
for each respective iteration:
locating a stored application in a catalog of applications previously used in the network, wherein:
the stored application is located in the catalog based on determining that-one of the one or more sub-graphs of the stored application is a closest matching sub-graph to its respective iteratively identified sub-graph of the logical-physical mapping of the application being added to the network;
running a neural network model by a computing device, by performing an experiment using the respective closest matching sub-graph of the one of the one or more sub-graphs of the stored application, wherein the experiment includes inducing a physical node mapping of respective closest matching sub-graph and determining a computing cost function; and
deploying the application being added to the network to a selected one or more physical nodes of one of the one or more respective induced physical node mapping of one of the closest matching sub-graphs based on determining a most effective computing cost function by comparing the computing cost function of each iteratively performed experiments.