Method for selecting a rank ordered sequence based on probabilistic dissimilarity matrix

• US 20070239652A1
• Filed: 03/09/2006
• Published: 10/11/2007
• Est. Priority Date: 03/09/2006
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Tuv, Eugene, Borisov, Alexander, Shahapurkar, Somnath
• Granted Patent: US 7,930,266 B2
• Time in Patent Office: Days
• US Class Current: 706/62
• CPC Class Codes: G06F 16/2465 Query processing support for facilitating data mining operations in structured databases
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A computer implemented method for ordering a plurality of entities by computing a dissimilarity matrix based on a plurality of probabilities. The pluralities of probabilities are determined based on a plurality of classes. A weighted distance matrix is computed based the dissimilarity matrix. A plurality of rank ordered sequence candidates based at least in part on the sum of weighted distances between neighboring entities in the rank ordered sequence is calculated. Other embodiments are described in the claims.

Claims

1. A computer implemented method for ordering a plurality of entities comprising:

computing a dissimilarity matrix based on a plurality of probabilities of classifications, wherein the classifications are determined by a quality of each entity;

computing a weighted distance matrix based on the dissimilarity matrix and weighted distances between at least one pair of the entities; and

selecting, from a plurality of rank ordered sequence candidates, a rank ordered sequence based at least in part on the sum of weighted distances between neighboring entities in the rank ordered sequence.

2. The method of claim 1, wherein the plurality of entities are microelectronic die.

3. The method of claim 1, wherein the sum of weighted distances is the smallest.

4. The method of claim 1, wherein the quality of each entity comprising chemical usage, wafer sorting, clocking speed, power consumption and electrical performance binning.

5. The method of claim 1, wherein the classification is die level cherry picking.

6. The method of claim 1, wherein the dissimilarity matrix includes dissimilarity elements s_km, such that

7. The method of claim 4, wherein the weighted distance matrix includes elements d_km, such that

8. The method of claim 1 further comprising applying one dimensional scaling to the weighted distance matrix to obtain the plurality of rank ordered sequence candidates.

9. A computer implemented method for ordering a plurality of microelectronic die comprising:

determining, based on historical data, a plurality of probabilities of classifications that each of the microelectronic die is in each of a plurality of classes, wherein the classifications are determined by a quality of each microelectronic die;

computing a dissimilarity matrix based on a plurality of probabilities of classifications,

computing a plurality of probabilities that a plurality of microelectronic die are in each of the plurality of classes;

computing a weighted distance matrix based on a weight, a geometric distance between a first microelectronic die and a second microelectronic die, and the dissimilarity matrix; and

retrieving the plurality of microelectronic die from a wafer area in an order based on a rank ordered sequence based at least in part on the sum of weighted distances between neighboring microelectronic die in the rank ordered sequence.

10. The method of claim 9, wherein the quality of each microelectronic die comprising a characteristic of the die used in microelectronic manufacturing process.

11. The method of claim 9, wherein the classification is die level cherry picking.

12. The method of claim 9, wherein the dissimilarity matrix includes dissimilarity elements s_km, such that

13. The method of claim 12, wherein the weighted distance matrix includes elements d_km , such that

14. The method of claim 9 further comprising applying one dimensional scaling to the weighted distance matrix to obtain the plurality of rank ordered sequence candidates.

15. A computer implemented system for ordering a plurality of entity comprising:

an artificial intelligence based module to determine, based on historical data, a plurality of probabilities of classifications that each of the entities is in each of a plurality of classes, wherein the classifications are determined by the quality of each microelectronic die.

a first data structure to represent the dissimilarity between a first entity and a second entity based on the determined probability;

a second data structure to represent the weighted distance between the first entity and the second entity based on a weight, and the geometric distance and the dissimilarity between the first entity and the second entity; and

an optimization module to select, from a plurality of rank ordered sequence candidates, a rank ordered sequence based at least in part on the sum of weighted distances between neighboring entities in the rank ordered sequence.

16. The system of claim 15, wherein sum of weighted distances is the smallest.

17. The system of claim 15, wherein the plurality of entities are microelectronic die.

18. The system of claim 15, wherein the artificial intelligence based module includes a random subspace classifier.

19. The system of claim 15, wherein the classification is die level cherry picking.

20. The system of claim 15, wherein dissimilarity between the first entity and the second entity is determined according to the formula:

21. The system of claim 20, wherein the weighted distance between the first entity and the second entity is determined according to the formula:

22. A machine-accessible medium that provides instructions that, when executed by a processor, causes the processor to:

construct a dissimilarity matrix based on a plurality of probability of classifications and a plurality of entities, wherein the classifications are determined based on the quality of the plurality of entities;

construct a weighted distance matrix based on a weight, a geometric distance between a first entity and a second entity, and the dissimilarity matrix; and

determine a rank ordered sequence from a plurality of rank ordered sequence candidates, a rank ordered sequence based at least in part on the sum of weighted distances between neighboring entities in the rank ordered sequence.

23. The machine-accessible medium of claim 22, wherein the sum of weighted distances is the smallest.

24. The machine-accessible medium of claim 22, wherein the classification is die-level cherry picking.

25. The machine-accessible medium of claim 22, wherein the dissimilarity matrix includes dissimilarity elements s_km, such that

26. The machine-accessible medium of claim 25, wherein the weighted distance matrix includes elements d_km such that

27. The machine-accessible medium of claim 22 further comprising applying one dimensional scaling to the weighted distance matrix to obtain the plurality of rank ordered sequence candidates.