COMPUTER BASED REASONING AND ARTIFICIAL INTELLIGENCE SYSTEMS

• US 20190310592A1
• Filed: 04/09/2018
• Published: 10/10/2019
• Est. Priority Date: 04/09/2018
• Status: Abandoned Application

• Current Assignee: Howso Incorporated
• Original Assignee: Diveplane Corporation
• Inventors: HAZARD, CHRISTOPHER JAMES
• Application Number:
  US15/948,805
• Publication Number:
  US 20190310592A1
• Time in Patent Office: Days
• CPC Class Codes: G05B 13/029 using neural networks and expert systems
• CPC Class Codes: G05B 13/04 involving the use of models or simulators
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 20/13 Satellite images
• CPC Class Codes: G06V 20/56 exterior to a vehicle by using sensors mounted on the vehicle

Abstract

Techniques are provided herein for creating well-balanced computer-based reasoning systems and using those to control systems. The techniques include receiving a request to determine whether to include or select one or more aspects in a computer-based reasoning model and determining two probability density or mass functions (“PDMFs”), one for the data set including the one or more particular aspects, once for the data set excluding it. Surprisal is determined based on those two PDMFs, and inclusion or selection in the computer-based reasoning model is determined based on surprisal. A system is later controlled using the computer-based reasoning model.

Claims

1. A method comprising:

receiving, at a training and analysis system, a request to determine whether to use one or more particular aspects of a computer-based reasoning model, wherein the training and analysis system executes on one or more computing devices, and is configured to execute training and analysis instructions, wherein the one or more particular aspects of the computer-based reasoning models include one or more features of data elements or one or more parameters of the computer-based reasoning model;

determining, at the training and analysis system, a first PDMF for a set of data elements associated with the computer-based reasoning model where the one or more particular aspects are excluded from the set of data elements and the computer-based reasoning model;

determining, at the training and analysis system, a second PDMF for the set of data elements with the one or more particular aspects included in the determination of the second PDMF,

determining, at the training and analysis system, surprisal of the one or more particular aspects based on a ratio of the first PDMF and the second PDMF;

in response to determining that the surprisal of the one or more particular aspects is out of bounds of one or more predetermined thresholds, including the one or more particular aspects in the computer-based reasoning model;

in response to determining that the surprisal of the one or more particular aspects is not out of bounds of the one or more predetermined thresholds, excluding the one or more particular aspects from the computer-based reasoning model;

causing, with a control system, control of a system with the computer-based reasoning model,

wherein the method is performed on one or more computing devices.

2. The method of claim 1, further comprising:

wherein the one or more particular aspects relate to at least one label on at least one training image;

wherein causing control of the system comprises causing control of a system that identifies elements of an image using the computer-based reasoning model by;

receiving an input image for labelling;

determining one or more labels for the input image based on the image and the computer-based reasoning model;

labelling the input image based on the one or more determined labels.

3. The method of claim 1, further comprising:

wherein causing control of the system comprises causing control of a vehicle using the computer-based reasoning model by;

receiving a current context for the vehicle, wherein the vehicle can be controlled by the control system;

determining an action to take for the vehicle based on the current context for the vehicle and the computer-based reasoning model;

causing the vehicle to perform the determined action.

4. The method of claim 1, wherein the one or more particular aspects relate to context-action pairs and the set of data elements includes a set of context action pairs and the first and second PDMFs are measures of probability density for context-action.

5. The method of claim 1, further comprising:

receiving a request for model reduction of the computer-based reasoning model;

determining surprisal of each aspect in the one or more particular aspects relative to the computer-based reasoning model without the aspect;

determining a subset of the one or more particular aspects based at least in part on the surprisal of each aspect in the one or more particular aspects, wherein the subset contains only aspects from the one or more particular aspects for which surprisal is not out of bounds of one or more predetermined thresholds; and

responding to the request for reduction of the with the one or more particular aspects.

6. The method of claim 1, further comprising:

initially receiving the one or more particular aspects as part of training for the computer-based reasoning model;

in response to determining that the surprisal of the one or more particular aspects is above a predetermined threshold, sending an indication to continue to collect the one or more particular aspects;

in response to determining that the surprisal of the one or more particular aspects is not above a predetermined threshold, sending the to no longer collect the one or more particular aspects.

7. The method of claim 1, further comprising

receiving a request for an action to take in a current context;

determining the action to take based on comparing the current context to contexts associated with data elements in the set of data elements; and

responding to the request for the action to take with the determined action.

8. The method of claim 1, further comprising determining the first PDMF using a parametric distribution.

9. The method of claim 1, further comprising determining the first PDMF using a nonparametric distribution.

10. The method of claim 1, further comprising:

determining multiple nearest data elements from the set of data elements for one or more particular data elements in the set of data elements;

determining multiple premetric contributions, one for each of the multiple nearest data elements;

determining a premetric measurement of the one or more particular data elements based at least in part on the multiple premetric contributions; and

determining new premetric measurements for at least one data element in the set of data elements, wherein each new premetric measurement for the at least one data element is computed based on premetric measurement to the particular data elements;

determining the second PDMF based at least in part on the premetric measurement for the one or more particular data elements and the new premetric measurements for the at least one data element in the set of data elements.

11. The method of claim 10, wherein:

the first PDMF is computed based on an average premetric contribution of each data element in the set of data elements divided by a sum of premetric contributions of each data element in the set of data elements;

and further comprising;

determining the second PDMF based on the premetric measurement of the one or more particular data elements divided by a sum of the new premetric measurements.

12. A non-transitory computer readable medium storing instructions which, when executed by one or more computing devices, cause the one or more computing devices to perform the method of claim 1.

13. A system for creating a computer-based reasoning model, comprising:

a training and analysis system executing on one or more computing devices, and configured to execute training and analysis instructions, which, when executed, perform the steps of;

receiving a request to determine whether to use one or more particular aspects of a computer-based reasoning model, wherein the one or more particular aspects of the computer-based reasoning models include one or more features of data elements or one or more parameters of the computer-based reasoning model;

determining a first PDMF for a set of data elements associated with the computer-based reasoning model where the one or more particular aspects are excluded from the set of data elements and the computer-based reasoning model;

determining a second PDMF for the set of data elements with the one or more particular aspects included in the determination of the second PDMF;

determining, at the training and analysis system, surprisal of the one or more particular aspects based on a ratio of the first PDMF and the second PDMF;

in response to determining that the surprisal of the one or more particular aspects is out of bounds of one or more predetermined thresholds, including the one or more particular aspects in the computer-based reasoning model;

in response to determining that the surprisal of the one or more particular aspects is not out of bounds of the one or more predetermined thresholds, excluding the one or more particular aspects from the computer-based reasoning model;

sending the computer-based reasoning model to a control system;

a control system executing on the one or more computing devices, configured to communicate with the training and analysis system and to execute control system instructions, which, when executed, perform the steps of;

receiving the computer-based reasoning model from the training and analysis system;

receiving a current context for a target system, wherein the target system can be controlled by the control system;

determining an action to take for the target system based on the current context for the target system and the computer-based reasoning model;

causing the target system to perform the determined action.

14. The system of claim 13, the training and analysis system further configured to perform the steps of:

determining multiple nearest data elements from the set of data elements for one or more particular data elements;

determining multiple premetric contributions, one for each of the multiple nearest data elements;

determining a premetric measurement of the one or more particular data elements based at least in part on the multiple premetric contributions; and

determining new premetric measurements for at least one data element in the set of data elements, wherein each new premetric measurement for the at least one data element is computed based on premetric measurement to the one or more particular data elements;

determining the second PDMF based at least in part on the premetric measurement for the one or more particular data elements and the new premetric measurements for the at least one data element in the set of data elements.

15. The system of claim 14, wherein:

the first PDMF is computed based on an average premetric contribution of each data element in the set of data elements divided by a sum of premetric contributions of each data element in the set of data elements;

and further comprising;

determining the second PDMF based on the premetric measurement of the one or more particular data elements divided by a sum of the new premetric measurements.

16. A system for creating a computer-based reasoning model for controlling vehicles, comprising:

a vehicular training and analysis system executing on one or more computing devices, and configured to execute training and analysis instructions, which, when executed, perform the steps of;

receiving a request to determine whether to use one or more particular aspects of a vehicular computer-based reasoning model, wherein the one or more particular aspects of the vehicular computer-based reasoning models include one or more features of vehicular data elements or one or more parameters of the vehicular computer-based reasoning model;

determining a first PDMF for a set of vehicular data elements associated with the vehicular computer-based reasoning model where the one or more particular aspects are excluded from the set of vehicular data elements and the vehicular computer-based reasoning model;

determining a second PDMF for the set of vehicular data elements with the one or more particular aspects included in the determination of the second PDMF;

determining, at the training and analysis system, surprisal of the one or more particular aspects based on a ratio of the first PDMF and the second PDMF;

in response to determining that the surprisal of the one or more particular aspects is out of bounds of one or more predetermined thresholds, including the one or more particular aspects in the vehicular computer-based reasoning model;

in response to determining that the surprisal of the one or more particular aspects is not out of bounds of one or more predetermined thresholds, excluding the one or more particular aspects from the vehicular computer-based reasoning model;

sending the vehicular computer-based reasoning model to a vehicular control system;

a vehicular control system executing on the one or more computing devices, configured to communicate with the vehicular training and analysis system and to execute control system instructions, which, when executed, perform the steps of;

receiving the vehicular computer-based reasoning model from the vehicular training and analysis system;

receiving a current vehicular context for a target vehicular system, wherein the vehicular target system can be controlled by the vehicular control system;

determining an action to take for the target system based on the current vehicular context for the target vehicular system and the vehicular computer-based reasoning model;

causing the target vehicular system to perform the determined action.

17. The system of claim 16, the vehicular training and analysis system further configured to perform the steps of:

determining multiple nearest vehicular data elements from the set of vehicular data elements;

determining multiple premetric contributions, one for each of the multiple nearest vehicular data elements;

determining a premetric measurement of the one or more particular vehicular data elements based at least in part on the multiple premetric contributions; and

determining new premetric measurements for at least one vehicular data element in the set of vehicular data elements, wherein each new premetric measurement for the at least one vehicular data element is computed based on premetric measurement to the one or more particular vehicular data elements;

determining the second PDMF based at least in part on the premetric measurement for the one or more particular vehicular data elements and the new premetric measurements for the at least one vehicular data element in the set of vehicular data elements.

18. The system of claim 17, wherein:

the first PDMF is computed based on an average premetric contribution of each vehicular data element in the set of vehicular data elements divided by a sum of premetric contributions of each vehicular data element in the set of vehicular data elements;

and further comprising;

determining the second PDMF based on the premetric measurement of each vehicular data element in the set of vehicular data elements divided by a sum of the new premetric measurements.

19. The system of claim 16, the steps further comprising determining the first PDMF using a parametric distribution.

20. The system of claim 16, the steps further comprising determining the first PDMF using a nonparametric distribution.