Machine learning techniques using cross-model fingerprints for novel predictive tasks

Patent metadata

Field	Value
Publication number	US-2023186151-A1
Application number	US-202117643921-A
Country	US
Kind code	A1
Filing date	Dec 13, 2021
Priority date	Dec 13, 2021
Publication date	Jun 15, 2023
Grant date	—

Abstract

Various embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and/or the like for performing predictive data analysis. Certain embodiments of the present invention utilize systems, methods, and computer program products that perform predictive data analysis by using affirmative fingerprint distance measures and negative fingerprint distance measures.

First claim

1 . A computer-implemented method for determining a predictive output for a predictive input in relation to novel predictive task using a plurality of existing task prediction machine learning models associated with a plurality of existing predictive tasks, the computer-implemented method comprising:
for each existing task prediction machine learning model, determining, based at least in part on the predictive input, and using one or more processors and the existing task prediction machine learning model, a per-model inferred representation for the predictive input; determining, using the one or more processors, a cross-model inferred representation for the predictive input based at least in part on each per-model inferred representation; determining, using the one or more processors and based at least in part on the cross-model inferred representation and an affirmative cross-model fingerprint for the novel predictive task, an affirmative fingerprint distance measure for the predictive input, wherein the affirmative fingerprint distance measure is determined by: (i) for each affirmative-labeled predictive input of one or more affirmative-labeled predictive inputs associated with the novel predictive task, processing the affirmative-labeled predictive input using the plurality of existing task prediction machine learning models to generate a plurality of affirmative-labeled per-model inferred representations for the affirmative-labeled predictive input, and (ii) determining the affirmative fingerprint distance measure based at least in part on each plurality of affirmative-labeled per-model inferred representations; determining, using the one or more processors, the predictive output based at least in part on the affirmative fingerprint distance measure; and performing, using the one or more processors, one or more prediction-based actions based at least in part on the predictive output.

2 . The computer-implemented method of claim 1 , wherein determining the predictive output further comprises:
determining, based at least in part on the cross-model inferred representation and a negative cross-model fingerprint for the novel predictive task, a negative fingerprint distance measure for the predictive input, wherein the negative fingerprint distance measure is determined by: (i) for each negative-labeled predictive input of one or more negative-labeled predictive inputs associated with the novel predictive task, processing the negative-labeled predictive input using the plurality of existing task prediction machine learning models to generate a plurality of negative-labeled per-model inferred representations for the negative-labeled predictive input, and (ii) determining the negative fingerprint distance measure based at least in part on each plurality of negative-labeled per-model inferred representations; and determining the predictive output based at least in part on the affirmative fingerprint distance measure and the negative fingerprint distance measure.

3 . The computer-implemented method of claim 2 , wherein determining the predictive output based at least in part on the affirmative fingerprint distance measure and the negative fingerprint distance measure further comprises:
determining, using a novel task prediction machine learning model and based at least in part on the affirmative fingerprint distance measure and the negative fingerprint distance measure, the predictive output.

4 . The computer-implemented method of claim 1 , wherein each per-model inferred representation for a particular existing task prediction machine learning model is determined based at least in part on one or more final output values generated by the particular existing task prediction machine learning model via processing the predictive input.

5 . The computer-implemented method of claim 1 , wherein each per-model inferred representation for a particular existing task prediction machine learning model is determined based at least in part on one or more intermediate output values generated by the particular existing task prediction machine learning model via processing the predictive input.

6 . The computer-implemented method of claim 5 , wherein the one or more intermediate output values generated by the particular existing task prediction machine learning model comprises one or more activation output values generated by the particular existing task prediction machine learning model via processing the predictive input.

7 . The computer-implemented method of claim 5 , wherein the one or more intermediate output values generated by the particular existing task prediction machine learning model comprises one or more heat map values generated by the particular existing task prediction machine learning model via processing the predictive input.

8 . The computer-implemented method of claim 1 , wherein determining the predictive output based at least in part on the affirmative fingerprint distance measure comprises:
determining, using a novel task prediction machine learning model and based at least in part on the affirmative fingerprint distance measure, the predictive output.

9 . The computer-implemented method of claim 1 , wherein the novel predictive task comprises a rare disease detection predictive task associated with a rare disease identifier.

10 . The computer-implemented method of claim 9 , wherein each existing task prediction machine learning model comprises a disease detection predictive task associated with a disease identifier that is distinct from the rare disease identifier.

11 . An apparatus for determining a predictive output for a predictive input in relation to novel predictive task using a plurality of existing task prediction machine learning models associated with a plurality of existing predictive tasks, the apparatus comprising at least one processor and at least one memory including program code, the at least one memory and the program code configured to, with the at least one processor, cause the apparatus to at least:
for each existing task prediction machine learning model, determine, based at least in part on the predictive input and using the existing task prediction machine learning model, a per-model inferred representation for the predictive input; determine a cross-model inferred representation for the predictive input based at least in part on each per-model inferred representation; determine, based at least in part on the cross-model inferred representation and an affirmative cross-model fingerprint for the novel predictive task, an affirmative fingerprint distance measure for the predictive input, wherein the affirmative fingerprint distance measure is determined by: (i) for each affirmative-labeled predictive input of one or more affirmative-labeled predictive inputs associated with the novel predictive task, processing the affirmative-labeled predictive input using the plurality of existing task prediction machine learning models to generate a plurality of affirmative-labeled per-model inferred representations for the affirmative-labeled predictive input, and (ii) determining the affirmative fingerprint distance measure based at least in part on each plurality of affirmative-labeled per-model inferred representations; determine the predictive output based at least in part on the affirmative fingerprint distance measure; and perform one or more prediction-based actions based at least in part on the predictive output.

12 . The apparatus of claim 11 , wherein determining the predictive output further comprises:
determining, based at least in part on the cross-model inferred representation and a negative cross-model fingerprint for the novel predictive ta

Assignees

• Optum Services Ireland Ltd

Inventors

• Selim Ahmed
• Godden Paul J
• Boss Gregory J
• Satterwhite Erin A
• Mendelsohn Nancy Joan
• Majerus Melanie

Classifications

• G06N5/02

  Knowledge representation; Symbolic representation · CPC title

• G06N20/00 Primary

  Machine learning · CPC title