US 12,314,839 B1
System and method for federated two-stage compression with federated joint learning
Zhu Li, Overland Park, KS (US); Paras Maharjan, Kansas City, MO (US); and Brian Galvin, Silverdale, WA (US)
Assigned to ATOMBEAM TECHNOLOGIES INC, Moraga, CA (US)
Filed by AtomBeam Technologies Inc., Moraga, CA (US)
Filed on Sep. 26, 2024, as Appl. No. 18/898,608.
Application 18/898,608 is a continuation in part of application No. 18/890,748, filed on Sep. 19, 2024.
Application 18/890,748 is a continuation in part of application No. 18/623,018, filed on Mar. 31, 2024, granted, now 12,119,848.
This patent is subject to a terminal disclaimer.
Int. Cl. H03M 7/00 (2006.01); G06N 3/0455 (2023.01); H03M 7/30 (2006.01)
CPC G06N 3/0455 (2023.01) [H03M 7/3082 (2013.01)] 9 Claims
OG exemplary drawing
 
1. A system for federated two-stage compression with federated joint learning, comprising one or more computers with executable instructions that, when executed, cause the system to:
process input data through a Variational Autoencoder with Vector Quantization on an edge server to generate a plurality of compressed data;
transmit the plurality of compressed data to a midserver, wherein the midserver converts the plurality of compressed data into a plurality of codewords using a codebook;
transmit the plurality of codewords to a centralized server where the plurality of codewords are converted to a plurality of universal codewords using a universal codebook;
train a large codeword model core using the plurality of universal codewords; and
deploy a trained large codeword model core wherein the large codeword model core receives a plurality of input data and generates a plurality of compressed outputs.