Contextual similarity measures for objects and retrieval, classification, and clustering using same
View Patent ↗A method of comparing first and second objects in a context comprises: maximizing or minimizing a quantitative comparison of the first object and a mixture of the second object and the context respective to a weighting parameter that controls the mixture; and outputting a comparison value based on the value of the weighting parameter determined by the maximizing or minimizing.
1. A storage medium storing instructions executable to compare first and second objects in a context by:
maximizing or minimizing a comparison function including at least one of ƒ(q,ω·p+(1−ω)·u) and ƒ(q,(1−ω)·p+ω·u) respective to a weighting parameter ω where the first object is represented as q, the second object is represented as p, the context is represented as u, and function ƒ is a similarity, difference, or divergence measure, and
outputting a comparison value that is equal to or determined based on a value of the weighting parameter ω obtained by the maximizing or minimizing.
2. The storage medium as set forth in claim 1 , wherein the comparison function is a similarity measure, and the instructions are executable to compare the first and second objects in the context by maximizing the similarity measure respective to the weighting parameter.
3. The storage medium as set forth in claim 2 , wherein the outputting comprises outputting a contextual similarity measure equal to the value of the weighting parameter that maximizes the similarity measure.
4. The storage medium as set forth in claim 1 , wherein the comparison function is a difference or divergence measure.
5. The storage medium as set forth in claim 1 , wherein the comparison function is selected from the class of Bregman divergences.
6. The storage medium as set forth in claim 1 , wherein the comparison function is selected from the class of Csizár divergences.
7. The storage medium as set forth in claim 1 , wherein the maximizing or minimizing of the comparison function respective to a weighting parameter ω is performed under the constraint that the weighting ω lies in an interval [0,1].
8. The storage medium as set forth in claim 1 , wherein the maximizing or minimizing of the comparison function respective to a weighting parameter is terminated responsive to the weighting parameter satisfying a termination criterion.
9. The storage medium as set forth in claim 1 , further storing instructions executable to perform a retrieval method including retrieving the first object conditional upon the outputted comparison value.
10. The storage medium as set forth in claim 1 , further storing instructions executable to perform a clustering method including grouping together first and second objects based at least in part on the outputted comparison value.
11. The storage medium as set forth in claim 1 , wherein the comparison function is symmetrized by further including a quantitative comparison of the second object p and a mixture of the first object q and the context u wherein the quantitative comparison of the second object p and the mixture of the first object q and the context u is also controlled by the weighting parameter ω.
12. The storage medium as set forth in claim 1 , wherein the comparison function comprises a symmetrized similarity measure scs ƒ (q,p)=arg max ω {ƒ(q,ω·p+(1−ω)·u)+ƒ(p,ω·q+(1−ω)·u)} where 0≦ω≦1.
13. A storage medium storing instructions executable to compare first and second objects in a context by:
maximizing or minimizing a comparison function respective to a weighting parameter, the comparison function including a quantitative comparison of the first object and a mixture of the second object and the context wherein the mixture is controlled by the weighting parameter, wherein the maximizing or minimizing comprises maximizing or minimizing ƒ(q,ω·p+(1−ω)·u) or ƒ(q,(1−ω)·p+ω·u) respective to the weighting parameter represented as ω, where the first object is represented as q, the second object is represented as p, the context is represented as u, and ƒ is a similarity, difference, or divergence measure; and
outputting a comparison value that is equal to or determined based on a value of the weighting parameter obtained by the maximizing or minimizing.
14. The storage medium as set forth in claim 13 , wherein the maximizing or minimizing is performed under the constraint 0≦ω≦1.
15. A method comprising:
comparing first and second objects in a context by maximizing or minimizing a comparison function including at least one of ƒ(q,ω·p+(1−ω)·u) and ƒ(q,(1−ω)·p+ω·u) respective to a weighting parameter ω where the first object is represented as q, the second object is represented as p, the context is represented as u, and function ƒ is a similarity, difference, or divergence measure; and
outputting a comparison value that is equal to or determined based on a value of the weighting parameter ω obtained by the maximizing or minimizing;
wherein the comparing and outputting operations are performed by a digital processor.
16. The method as set forth in claim 15 , further comprising:
retrieving the first object q conditional upon the outputted comparison value satisfying a threshold.
17. The method as set forth in claim 15 , wherein the outputting comprises:
outputting a binary decision value indicative of whether the value of the weighting parameter ω determined by the maximizing or minimizing satisfies a threshold.
18. A digital processor configured to perform the method of claim 15 .