Task-based design evaluation

• US 20050261881A1
• Filed: 05/21/2004
• Published: 11/24/2005
• Est. Priority Date: 05/21/2004
• Status: Active Grant

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Jackson, Warren Bruce
• Granted Patent: US 9,235,655 B2
• Time in Patent Office: Days
• US Class Current: 703/1
• CPC Class Codes: G06F 2111/02 CAD in a network environment, e.g. collaborative CAD or distributed simulation
• CPC Class Codes: G06F 30/00 Computer-aided design [CAD]
• CPC Class Codes: G06F 30/10 Geometric CAD

Abstract

A proposed feature vector for a first deign is received and an existing feature vector for an existing design is retrieved for a given task. The proposed design is evaluated against the existing design using task-based scores associated with each design and based on their performances for the given task.

Claims

1. A method, comprising:

selecting a task to evaluate a proposed design against an existing design;

receiving a proposed feature vector for the proposed design;

retrieving an existing feature vector for the existing design; and

evaluating if the proposed design has a more desirable estimated score than an existing score for the existing design based on the proposed feature vector and the existing feature vector for the selected task.

2. The method of claim 1 further comprising, selecting the task from a knowledge store as an off-line task.

3. The method of claim 1 further comprising, acquiring the task interactively as an on-line task.

4. The method of claim 1 further comprising, generating a new design if the estimated score is less than existing score, wherein the new design includes a new feature vector and a new score that is higher than the existing score or the proposed score.

5. The method of claim 1, wherein evaluating further includes mapping the proposed feature vector to the estimated score based on a mapping function, wherein the mapping function also produces the existing score for the existing feature vector.

6. The method of claim 5, wherein mapping further includes processing the mapping function by executing at least one of a weighting algorithm against individual proposed feature elements in the proposed feature vector and summing a result to acquire the estimated score, a summing algorithm that sums individual scores assigned to individual proposed feature elements to acquire the estimated score, and a neural network algorithm that maps the proposed feature vector to the estimated score.

7. The method of claim 1 further comprising, generating the proposed feature vector for the proposed design by evaluating a received document that implements the proposed design.

8. The method of claim 1 further comprising, identifying differences between the proposed feature vector and the existing feature vector.

9. A method, comprising:

selecting a task to evaluate variations of a model design;

receiving a plurality of feature vectors, one feature vector representing the model design, and the remaining feature vectors representing the variations to the model design;

assigning scores to the feature vectors as a function of applying the task to the model design and the variations of the model design; and

learning a mapping between the scores and the feature vectors.

10. The method of claim 9, wherein learning further includes processing a support vector machine.

11. The method of claim 9, wherein receiving further includes processing a feature vector extractor and a design receiver.

12. The method of claim 9, wherein receiving further includes spanning a number of the feature vectors within a feature vector space.

13. The method of claim 9, wherein assigning further includes representing the task as a low level perceptual task.

14. The method of claim 9, wherein assigning further includes representing the task as a just noticeable difference task.

15. The method of claim 9, wherein assigning further includes determining the score based on weights obtained from the task which is a just noticeable difference task.

16. The method of claim 9, wherein selecting further includes acquiring the task as an off-line task or an on-line task.

17. A method, comprising:

acquiring a task to evaluate a design;

acquiring the design and one or more additional designs;

generating features for the design and the one or more additional designs;

scoring the design and the one or more additional designs using the task; and

evaluating correlation strengths between variations in the features against variations in scores for the design and the one or more additional designs.

18. The method of claim 17, wherein evaluating further includes using a principal component analysis.

19. The method of claim 17, wherein acquiring the task further includes interactively acquiring the task as an on-line task.

20. The method of claim 17, wherein acquiring the task further includes acquiring the task from a knowledge store as an off-line task.

21. The method of claim 17, wherein acquiring the task further includes acquiring the task as a low-level vision discernment task.

22. A data structure, implemented in a computer readable medium, for evaluating designs, the data structure comprising:

a plurality of features defining aspects of a design;

a feature vector string representing a concatenation of the plurality of vectors; and

a task-based score for the feature vector string which is dynamically resolved by processing a mapping function for a selected task.

23. The data structure of claim 22, wherein each of the features identify a particular element or element attribute included within the design.

24. The data structure of claim 22, wherein each feature identifies a structural placement for a particular element within a document that conforms to the design.

25. The data structure of claim 22, wherein a number of the features define semantics associated with values that particular elements can have within a document that conforms to the design.

26. The data structure of claim 22 further comprising, a reference to a task determining application that permits the task to be determined or selected.

27. The data structure of claim 22, further comprising a unique identifier for the data structure which is used for housing instances of the data structure in a data store.

28. The data structure of claim 22, wherein the data structure is dynamically instantiated by a parsing application that parses a document which implements the design.

29. The data structure of claim 22, wherein the selected task is acquired as an off-line task from a knowledge store or as an on-line task from interactive and dynamic processing.

30. A system, comprising:

first and second designs;

first and second feature vectors, wherein the first feature vector represents the first design and the second feature vector represents the second design; and

a design evaluator that evaluates a first task-based score associated with the first feature vector against a second task-based score associated with the second feature vector for a given selected task.

31. The system of claim 30 further comprising a task determiner that assists in selecting the task.

32. The system of claim 30 further comprising a mapper that interfaces with the design evaluator for mapping the first feature vector to the first task-based score and mapping the second feature vector to the second task-based score.

33. The system of claim 32 further comprising, a design generator that parses one or more documents to generate at least one of the designs.

34. The system of claim 33, wherein the design generator interfaces with the design evaluator to generate a third design which includes differences between the first and second feature vectors and at least one additional change, and wherein the third design is represented by a third feature vector and a third task-based score and the third task-based score is greater than the first or second task-based scores for the task.

35. The system of claim 30, wherein the third design is used as a template for producing one or more instances of a document.

36. The system of claim 30 further comprising a knowledge store that houses the task and a plurality of other tasks, and wherein the design evaluator is interfaced to the knowledge store.

37. The system of claim 30, wherein the given selected task is acquired as an off-line task or an on-line task.

38. A computer readable medium for evaluating designs having instructions thereon, the instructions when accessed performing the method comprising:

receiving first and second feature vectors associated with first and second designs;

identifying a task for measuring the performance of the first and second designs against one another;

mapping the second feature vector to a second task-based score based on the task; and

evaluating a first task-based score associated with the first feature vector against the second task-based score.

39. The medium of claim 38 further including instructions for selecting one of the designs as a desired design, if the selected design has a greater task-based score.

40. The medium of claim 38 further including instructions for generating a third design represented by a third feature vector, wherein the third feature vector includes changes associated with moving the first feature vector to the second feature vector and at least one additional change, and wherein the third feature vector includes a third task-based score that is greater than the first or second task-based scores.

41. The medium of claim 38, further including instructions for generating at least one of the first feature vector and first design from a first document, and the second feature vector and the second design from a second document.

42. The medium of claim 38, wherein the identifying instructions further include acquiring the task offline from a knowledge store or interactive online.