US 12,107,665 B2
Generating model update data at satellite
Tsu-wang Hsieh, Redmond, WA (US); Jin Hyun So, Los Angeles, CA (US); Behnaz Arzani, Redmond, WA (US); Shadi Abdollahian Noghabi, Bellevue, WA (US); and Ranveer Chandra, Kirkland, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Jan. 26, 2022, as Appl. No. 17/649,044.
Prior Publication US 2023/0239042 A1, Jul. 27, 2023
Int. Cl. H04B 7/185 (2006.01); H04B 7/15 (2006.01); H04B 17/391 (2015.01)
CPC H04B 7/18539 (2013.01) [H04B 7/18586 (2013.01); H04B 7/18589 (2013.01); H04B 17/391 (2015.01)] 18 Claims
OG exemplary drawing
 
1. A satellite comprising:
an onboard computing device including a processor configured to:
receive training data while the satellite is in orbit;
perform training of a neural network based at least in part on the training data to thereby generate a modified neural network with modified parameters;
generate neural network update data including the modified neural network parameters and/or a gradient with respect to the modified neural network parameters; and
transmit the neural network update data from the satellite to an additional computing device.