US 12,293,265 B1
Apparatus and method for model optimization
Barbara Sue Smith, Toronto (CA); and Daniel J. Sullivan, Toronto (CA)
Assigned to The Strategic Coach Inc., Toronto (CA)
Filed by The Strategic Coach Inc., Toronto (CA)
Filed on Jan. 10, 2024, as Appl. No. 18/409,666.
Int. Cl. G06N 20/00 (2019.01)
CPC G06N 20/00 (2019.01) 14 Claims
OG exemplary drawing
 
1. A method for identifying model optimization, the method comprising:
measuring, by a computing device, a plurality of subsystems, wherein the plurality of subsystems includes at least one remote device, and wherein measuring the plurality of subsystems produces a plurality of measurements;
comparing, by the computing device, each measurement of the plurality of measurements to a predetermined continuum range, wherein the predetermined continuum range comprises a lower threshold and an upper threshold;
identifying, by the computing device, an optimal measurement of the plurality of measurements and a suboptimal measurement of the plurality of measurements as a function of each of the comparisons;
generating, by the computing device, a positive feedback function of the optimal measurement, wherein generating the positive feedback function further comprises identifying, using an optimal machine-learning model, a first set of parameter changes to a subsystem corresponding to the optimal measurement, wherein the optimal machine-learning model is generated by updating a previous optimal machine learning model, wherein updating the previous optimal machine learning model comprises:
receiving optimal training data, wherein the optimal training data comprises a plurality of data entries containing a plurality of optimal measurements as inputs correlated to a plurality of positive feedbacks as outputs;
training the previous machine-learning model using the optimal training data;
sanitizing the optimal training data using a dedicated hardware unit comprising circuitry configured to perform signal processing operations, wherein sanitizing the optimal training data comprises:
determining by the dedicated hardware unit that a training data entry has a signal to noise ratio below a threshold value; and
removing the training data entry from the optimal training data;
retraining the previous machine-learning model using the sanitized training data; and
generating the positive feedback function as a function of the optimal measurement using the retrained optimal machine-learning model;
generating, by the computing device, a negative feedback function of the suboptimal measurement, wherein generating the negative feedback function comprises identifying, using a suboptimal machine-learning model, a second set of parameter changes to a subsystem corresponding to the suboptimal measurement; and
configuring the at least one remote device using the positive feedback function and the negative feedback function.