Detecting and correcting anomalies in computer-based reasoning systems

Patent metadata

Field	Value
Publication number	US-9922286-B1
Application number	US-201715681219-A
Country	US
Kind code	B1
Filing date	Aug 18, 2017
Priority date	Aug 18, 2017
Publication date	Mar 20, 2018
Grant date	Mar 20, 2018

Abstract

Techniques for detecting and correcting anomalies in computer-based reasoning systems are provided herein. The techniques can include obtaining current context data and determining a contextually-determined action based on the obtained context data and a reasoning model. The reasoning model may have been determined based on multiple sets of training data. The techniques may cause performance of the contextually-determined action and, potentially, receiving an indication that performing the contextually-determined action in the current context resulted in an anomaly. The techniques include determining a portion of the reasoning model that caused the determination of the contextually-determined action based on the obtained context data and causing removal of the portion of the model that caused the determination of the contextually-determined action, to produce a corrected reasoning model. Subsequently, second context data is obtained, a second action is determined based on that data and the corrected reasoning model, and the second contextually-determined action can be performed.

First claim

What is claimed is:

1. A method comprising:
obtaining, using one or more computing devices, context data for a current context of a self-driving vehicle;
determining, using the one or more computing devices, a contextually-determined action for the self-driving vehicle based on the obtained context data and a reasoning model, wherein the reasoning model was determined based on multiple sets of training data, wherein the multiple sets of training data include multiple context data and action data pairings, and wherein determining the contextually-determined action for the self-driving vehicle comprises determining, using a premetric, closest context data in the multiple sets of training data that is closest to the current context based on the premetric and determining an action paired with the closest context data as the contextually-determined action for the self-driving vehicle, wherein the premetric is a Minkowski distance measure of order zero;
determine, using the one or more computing devices, whether performance of the contextually-determined action results in an indication of an anomaly for the self-driving vehicle;
determining, using the one or more computing devices, a portion of the reasoning model that caused the determination of the contextually-determined action that resulted in the indication of the anomaly for the self-driving vehicle based on the obtained context data;
updating, using the one or more computing devices, the portion of the reasoning model that caused the determination of the contextually-determined action that resulted in the indication of the anomaly for the self-driving vehicle, in order to produce a corrected reasoning model;
obtaining, using the one or more computing devices, subsequent contextual data for a second context for the self-driving vehicle;
determining, using the one or more computing devices, a second contextually-determined action for the self-driving vehicle based on the obtained subsequent contextual data and the corrected reasoning model; and
causing performance, using the one or more computing devices, of the second contextually-determined action for the self-driving vehicle.

2. The method of claim 1 , wherein the reasoning model is a case-based reasoning model.

3. The method of claim 1 , wherein determining the portion of the reasoning model that caused the determining of the contextually-determined action that resulted in the indication of the anomaly comprises determining the previously-identified closest context data in the training data.

4. The method of claim 3 , wherein updating the portion of the reasoning model that caused the determining of the contextually-determined action that resulted in the indication of the anomaly to produce the corrected reasoning model comprises removing an association between the previously-identified closest context data and the action paired with the closest context data.

5. The method of claim 1 , wherein updating the portion of the reasoning model that caused the determining of the contextually-determined action that resulted in the indication of the anomaly to produce the corrected reasoning model comprises removing the closest context data and the paired action.

6. The method of claim 1 , wherein the method additionally comprises determining additional portions of the reasoning model that would cause the performance of the contextually-determined action that resulted in the indication of the anomaly in the current context and removing the additional portions of the reasoning model.

7. One or more non-transitory storage media storing instructions which, when executed by the one or more computing devices, cause performance of the method recited in claim 1 .

8. A system for performing a machine-executed operation involving instructions, wherein said instructions are instructions which, when executed by one or more computing devices, cause performance of certain steps including:
obtaining context data for a current context of a self-driving vehicle;
determining a contextually-determined action for the self-driving vehicle based on the obtained context data and a reasoning model, wherein the reasoning model was determined based on multiple sets of training data, wherein the multiple sets of training data include multiple context data and action data pairings, and wherein determining the contextually-determined action for the self-driving vehicle comprises determining, using a premetric, closest context data in the multiple sets of training data that is closest to the current context based on the premetric and determining an action paired with the closest context data as the contextually-determined action for the self-driving vehicle, wherein the premetric is a Minkowski distance measure of order zero;
determine whether performance of the contextually-determined action results in an indication of an anomaly for the self-driving vehicle;
determining a portion of the reasoning model that caused the determining of the contextually-determined action that resulted in the indication of the anomaly for the self-driving vehicle based on the obtained context data;
updating the portion of the reasoning model that caused the determining of the contextually-determined action that resulted in the indication of the anomaly for the self-driving vehicle based on the obtained context data in order to produce a corrected reasoning model;
obtaining subsequent contextual data for a second context for the self-driving vehicle;
determining a second contextually-determined action for the self-driving vehicle based on the obtained subsequent contextual data and the corrected reasoning model; and
causing performance of the second contextually-determined action for the self-driving vehicle.

9. The system of claim 8 , wherein determining the portion of the reasoning model that caused the determining of the contextually-determined action that resulted in the indication of the anomaly comprises determining the previously-identified closest context data in the training data.

10. The system of claim 8 , wherein updating the portion of the reasoning model that cause the determining of the contextually-determined action that resulted in the indication of the anomaly to produce the corrected reasoning model comprises removing the portion of the reasoning model associated the closest context data, and the action paired with the closest context data.

11. The system of claim 8 , wherein updating the portion of the reasoning model that cause the determining of the contextually-determined action that resulted in the indication of the anomaly to produce the corrected reasoning model comprises changing the action paired with the closest context data.

12. The system of claim 8 , wherein the performed steps additionally comprise determining additional portions of the reasoning model that would cause the performance of the contextually-determined action that resulted in the indication of the anomaly and removing the additional portions of the reasoning model.

13. The system of claim 8 , wherein the reasoning model is a case-based reasoning model.

Assignees

• Hazardous Software Inc
• Diveplane Corp

Inventors

• Hazard Christopher James

Classifications

• G06F17/30539

  Physics · mapped topic

• G06N5/025 Primary

  Extracting rules from data · CPC title

• G06F16/2457 Primary

  with adaptation to user needs · CPC title

• G06N5/041

  Abduction · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title