US 12,282,839 B2
Constraint based inference and machine learning system
Joshua G. Fadaie, Saint Louis, MO (US); and Richard Hanes, Ballwin, MO (US)
Assigned to The Boeing Company, Arlington, VA (US)
Filed by The Boeing Company, Chicago, IL (US)
Filed on Nov. 17, 2021, as Appl. No. 17/529,067.
Claims priority of provisional application 63/118,495, filed on Nov. 25, 2020.
Prior Publication US 2022/0164636 A1, May 26, 2022
Int. Cl. G06N 3/047 (2023.01); G05B 13/02 (2006.01); G06N 3/045 (2023.01)
CPC G06N 3/047 (2023.01) [G05B 13/027 (2013.01); G06N 3/045 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving a pre-determined constraint on a plurality of user actions;
generating a constraint vector based on the pre-determined constraint;
inputting the constraint vector into a machine learning model;
generating a first output from the machine learning model by executing the machine learning model using the constraint vector as a first input to the machine learning model;
converting the constraint vector into a legal action mask;
generating a probability vector by executing a masked softmax operator, wherein:
the masked softmax operator takes, as a second input, the first output,
the masked softmax operator takes, as a third input, the legal action mask, and
the masked softmax operator generates, as a second output, the probabilities vector; and
generating a plurality of action outputs by applying a sampling system to the probability vector, wherein the plurality of action outputs comprise a subset of the plurality user actions, and wherein the subset includes only allowed user actions.