Method and system for obtaining and analyzing information from a plurality of sources

• US 10,062,034 B2
• Filed: 06/02/2016
• Issued: 08/28/2018
• Est. Priority Date: 06/08/2015
• Status: Expired due to Fees

• Current Assignee: Charles Stark Draper Laboratory Incorporated
• Original Assignee: Charles Stark Draper Laboratory Incorporated
• Inventors: Irvine, John, Abramson, Mark, Tuohy, Seamus
• Primary Examiner(s): Ouellette, Jonathan P
• Application Number:
  US15/171,158
• Publication Number:
  US 20160358090A1
• Time in Patent Office: 817 Days
• Field of Search: 705 11-912
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/045 Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 50/26 Government or public services business processes related to the transportation industry G06Q50/40
• CPC Class Codes: G06Q 50/265 Personal security, identity or safety

Abstract

Methods and systems are disclosed for obtaining and analyzing information to determine an estimated state of a real world. First information is obtained from a first source of a plurality of sources (e.g., data stores, real-time data streams, or information services), where the first information is potentially indicative of first and second states of the real world. If the first information exceeds a first decision threshold, then the estimated state is the first state, and if the first information exceeds a second decision threshold, then the estimated state is the second state. If the first information does not exceed either of the first and second decision thresholds, then second information is obtained from a second source. Obtaining the second information is based on whether the second source includes information that is relevant to the estimated state and whether the second source includes information that is redundant with the first information.

Claims

1. A method of obtaining and analyzing information to determine an estimated state of a real world, the method comprising:

obtaining, by a processor in communication with a plurality of sources, first information from a first source of the plurality of sources, the first information being potentially indicative of a first state of the real world and a second state of the real world;

determining, by the processor, that the estimated state is the first state if the first information exceeds a first decision threshold;

determining, by the processor, that the estimated state is the second state if the first information exceeds a second decision threshold;

obtaining, by the processor, second information from a second source of the plurality of sources if the first information does not exceed the first decision threshold and does not exceed the second decision threshold, the obtaining of the second information including;

determining, by the processor, whether the second source includes information that is relevant to the estimated state based on a divergence between (i) a joint density of the second information and the estimated state and (ii) the product of the densities of the second information and the estimated state, the determined divergence resulting in a relevant score;

determining, by the processor, whether the second source includes information that is redundant with the first information based on a divergence between (i) a joint density of the second information and the first information and (ii) the product of the densities of the second information and the first information, the determined divergence resulting in a redundant score; and

determining, by the processor, whether the second source includes information that is more relevant and less redundant than other sources of the plurality of sources by (i) for each source, subtracting from the relevant score corresponding to the source, the redundant score corresponding to the source and a cost of obtaining the information from the source, resulting in an overall score for each source, and (ii) selecting the second source if the overall score for the second source is greater than each overall score of the other sources of the plurality of sources;

determining, by the processor, whether the estimated state is the first state or the second state based on whether the first information and the second information together exceeds the first decision threshold or the second decision threshold, improving performance of the processor by reducing time consumed for determining whether the estimated state is the first state or the second state; and

generating a notification based on the estimated state determination.

2. A method as in claim 1 wherein determining whether the second source includes information that is more relevant and less redundant than the other sources includes weighting the redundant scores and the costs of obtaining the information.

3. A method as in claim 1 wherein the plurality of sources include any of a data store, a real-time data stream, and a service from which information can be requested.

4. A method as in claim 1 wherein:

the real world includes a plurality of ships traveling in a body of water;

the estimated state is whether a particular ship of the plurality of ships is a drug smuggler;

the first state is that the ship is a drug smuggler;

the second state is that the ship is not a drug smuggler;

the first information includes information regarding the type of the particular ship; and

the second information includes information regarding the route of the particular ship.

5. A method as in claim 1 wherein the first source and the second source are physical sources and include any of a temperature sensor, electric meter, and physiological monitor.

6. A system for obtaining and analyzing information to determine an estimated state of a real world, the system comprising:

a plurality of sources including any of a data store, a real-time data stream, and a service from which information can be requested, the plurality of sources including (i) a first source including first information potentially indicative of a first state of the real world and a second state of the real world and (ii) a second source including second information potentially indicative of the first state and the second state;

an interface in communication with the plurality of sources and configured to obtain information from the plurality of sources; and

a processor in communication with the interface and configured to;

determine that the estimated state is the first state if the first information exceeds a first decision threshold;

determine that the estimated state is the second state if the first information exceeds a second decision threshold;

obtain the second information from the second source if the first information does not exceed the first decision threshold and does not exceed the second decision threshold, the processor obtaining the second information by;

determining whether the second source includes information that is relevant to the estimated state based on a divergence between (i) a joint density of the second information and the estimated state and (ii) the product of the densities of the second information and the estimated state, the determined divergence resulting in a relevant score;

determining whether the second source includes information that is redundant with the first information based on a divergence between (i) a joint density of the second information and the first information and (ii) the product of the densities of the second information and the first information, the determined divergence resulting in a redundant score; and

determining whether the second source includes information that is more relevant and less redundant than other sources by (i) for each source, subtracting from the relevant score corresponding to the source, the redundant score corresponding to the source and a cost of obtaining the information from the source, resulting in an overall score for each source, and (ii) selecting the second source if the overall score for the second source is greater than each overall score of the other sources of the plurality of sources;

determine whether the estimated state is the first state or the second state based on whether the first information and the second information together exceeds the first decision threshold or the second decision threshold, improving performance of the processor by reducing time consumed for determining whether the estimated state is the first state or the second state; and

generate a notification based on the estimated state determination.

7. A system as in claim 6 wherein the processor is configured to weight the redundant scores and the costs of obtaining the information.

8. A machine readable storage medium having stored thereon a computer program for obtaining and analyzing information to determine an estimated state of a real world, the computer program comprising a routine of set instructions for causing the machine to:

obtain first information from a first source of a plurality of sources in a database, the first information being potentially indicative of a first state of the real world and a second state of the real world;

determine that the estimated state is the first state if the first information exceeds a first decision threshold;

determine that the estimated state is the second state if the first information exceeds a second decision threshold;

obtain second information from a second of the plurality of sources if the first information does not exceed the first decision threshold and does not exceed the second decision threshold, the obtaining of the second information including;

determining whether the second source includes information that is relevant to the estimated state based on a divergence between (i) a joint density of the second information and the estimated state and (ii) the product of the densities of the second information and the estimated state, the determined divergence resulting in a relevant score;

determining whether the second source includes information that is redundant with the first information based on a divergence between (i) a joint density of the second information and the first information and (ii) the product of the densities of the second information and the first information, the determined divergence resulting in a redundant score; and

determining whether the second source includes information that is more relevant and less redundant than other sources of the plurality of sources by (i) for each source, subtracting from the relevant score corresponding to the source, the redundant score corresponding to the source and a cost of obtaining the information from the source, resulting in an overall score for each source, and (ii) selecting the second source if the overall score for the second source is greater than each overall score of the other sources of the plurality of sources;

determine whether the estimated state is the first state or the second state based on whether the first information and the second information together exceeds the first decision threshold or the second decision threshold, improving performance of the machine by reducing time consumed for determining whether the estimated state is the first state or the second state; and

generate a notification based on the estimated state determination.

9. A machine readable storage medium as in claim 8 wherein the instructions cause the machine to weight the redundant scores and the costs of obtaining the information.