Predictive hypothesis exploration using planning

• US 9,785,755 B2
• Filed: 05/21/2014
• Issued: 10/10/2017
• Est. Priority Date: 05/21/2014
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Riabov, Anton V., Sohrabi Araghi, Shirin, Udrea, Octavian
• Primary Examiner(s): Smith, Paulinho E
• Application Number:
  US14/283,867
• Publication Number:
  US 20150339580A1
• Time in Patent Office: 1,238 Days
• Field of Search: None
• CPC Class Codes: G06F 11/00 Error detection; Error correction; Monitoring error detection, correction or monitoring in information storage based on relative movement between record carrier and transducer G11B20/18; monitoring, i.e. supervising the progress of recording or reproducing G11B27/36; in static stores G11C29/00
• CPC Class Codes: G06F 21/00 Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

In at least one embodiment, a method and a system include receiving a trace into a hypotheses generator from a source a trace, translating the trace and a state transition model into a planning problem using the hypotheses generator, producing a set of plans for the trace using at least one planner, translating each plan into hypothesis using the hypotheses generator and/or the planner, and returning the hypotheses from the hypotheses generator. In a further embodiment, the trace includes at least one of a future observation and a past observation. In at least one embodiment, the system includes at least one planner that develops a set of plans, a hypothesis generator, a database, at least one analytic, and at least one sensor where the hypotheses generator and/or the at least one planner converts each plan into a respective hypothesis.

Claims

1. A method for operation of a system for predicting hypotheses, the hypotheses comprising explanations of observations, said method comprising:

receiving a trace from a source of a trace, where a hypotheses generator in the system receives the trace;

translating the trace and a state transition model into a planning problem, where the hypotheses generator translates the trace and the state transition model into the planning problem;

producing a set of plans for the trace, where a planner produces the set of plans, each plan including a sequence of action operators, each action operator maps a state into another state;

translating each plan into a hypothesis, where at least one of the hypotheses generator and the planner translates each plan into the respective hypothesis; and

returning the hypotheses, where the hypotheses generator returns the hypotheses, and

wherein at least one of the set of plans includes at least one top-quality plan and the hypotheses includes at least one plausible hypothesis.

2. The method according to claim 1, wherein returning the hypotheses includes

scoring with the at least one planner each plan based on the cost of the plan associated with the hypothesis;

and

ordering the hypotheses based on the scores of the respective plans.

3. The method according to claim 2, wherein action costs for less probable actions is higher than action costs for more probable actions.

4. The method according to claim 2, wherein actions associated with bad states have a higher action cost than good states.

5. The method according to claim 2, wherein malicious actions have a higher action cost than non-malicious actions.

6. The method according to claim 1, wherein returning the hypotheses includes displaying each hypothesis including any observations present in the trace and the states with linkages between a plurality of the observations and the states,

wherein the hypotheses generator displays the hypotheses on a display, and

wherein each hypothesis is a sequence of actions that explains observations in a system.

7. The method according to claim 1, further comprising at least one of

requesting additional observations;

and

alerting an individual of any potential problem determined based on the generated hypotheses where the top hypothesis is indicative of a problem.

8. A method for operation of a system for predicting hypotheses, the hypotheses comprising explanations of observations, said method comprising:

receiving a trace including a request for at least one future observation or at least one past observation from a source of a trace, where a hypotheses generator receives the trace;

translating the trace and a state transition model into a planning problem, where the hypotheses generator translates the trace and the state transition model into the planning problem;

producing a set of plans for the trace, where at least one planner produces the set of plans, each plan including a sequence of action operators, each action operator maps a state into another state;

translating each plan into a hypothesis, where at least one of the hypotheses generator and the at least one planner translates each plan, where at least one hypothesis includes at least one of a future state and a past state; and

returning the hypotheses from the hypotheses generator, and

wherein the set of plans includes at least one top-quality plan and the hypotheses includes at least one plausible hypothesis.

9. The method according to claim 8, wherein the cost of plans is based on the individual costs of the actions present in the plan.

10. The method according to claim 9, wherein the cost of the actions is determined in part based on whether the action is at least one of more plausible or a good action.

11. The method according to claim 8, wherein returning the hypotheses includes displaying each hypothesis by the hypotheses generator through a display including any observations present in the trace and the states with linkages between a plurality of the observations and the states.

12. The method according to claim 8, further comprising returning future observations for at least one hypothesis returned by the hypotheses generator.

13. A system comprising:

at least one planner for the development of a set of plans;

a hypothesis generator in communication with said at least one planner;

a database in communication with said hypothesis generator and said at least one planner;

at least one analytic in communication with said hypotheses generator and said database; and

at least one sensor in communication with one of said at least one analytic, and

wherein at least one of said hypotheses generator and said at least one planner converts each plan in the set of plans into a hypothesis, each plan in the set of plans including a sequence of action operators, each action operator maps a state into another state, the hypothesis comprising an explanation of observations.

14. The system according to claim 13, wherein said at least one analytic converts the data received from said at least one sensor into observations that are at least one of stored in said database and communicated with said hypotheses generator.

15. The system according to claim 14, wherein said hypotheses generator converts the observations and a transition state model into a planning problem that is provided to said at least one planner.

16. The system according to claim 15, wherein said hypotheses generator inserts at least one past observation or at least one future observation into the planning problem.

17. The system according to claim 13, wherein said at least one planner determines a plan cost for each plan based on a predetermined action costs for the actions present in the plan.

18. The system according to claim 17, wherein said hypotheses generator displays each of the resulting hypotheses with a respective score based on the plan cost.

19. A computer program product for predicting hypotheses, the hypotheses comprising explanations of observations, the computer program product comprising:

a non-transitory computer readable storage medium having encoded thereon;

first program instructions executable by a processor to cause the processor to receive a trace;

second program instructions executable by a processor to cause the processor to translate the trace and a state transition model into a planning problem;

third program instructions executable by a processor to cause the processor to produce a set of plans for the trace, each plan including a sequence of action operators, each action operator maps a state into another state;

fourth program instructions executable by a processor to cause the processor to translate each plan into a hypothesis;

fifth program instructions executable by a processor to cause the processor to return the hypotheses; and

wherein at least one of the set of plans include at least one top-quality plan and the hypotheses include at least one most plausible hypothesis.

20. The computer program product according to claim 19, wherein

the trace includes at least one future observation or at least one past observation;

at least one generated hypothesis includes at least one of a future state and one past state.