Loading...
Please wait, while we are loading the content...
Similar Documents
Structure Discovery from Partial Rankings
| Content Provider | Semantic Scholar |
|---|---|
| Author | Huang, Jonathan |
| Copyright Year | 2010 |
| Abstract | Aggregating and statistical reasoning with ranked data are tasks that arise in a number of applications from analyzing political elections to modeling user preferences over a set of items. Representing distributions over rankings, however, can be daunting due to the fact that the number of rankings of n items scales factorially. Moreover, it is crucial for probabilistic models over rankings to be able to handle partially ranked data since real world data more often consists of partial rankings rather than full. In this work, we study a class of models over rankings called hierarchical riffle independent models, which can be thought of as being analogous to graphical models but more appropriate for ranked data. We show in particular that Bayesian conditioning based on top-k partial ranking evidence can be performed efficiently in these models, and apply our algorithms to estimate the structure of a riffle independent model from top-k rankings. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://people.cs.umass.edu/~wallach/workshops/nips2010css/papers/huang.pdf |
| Alternate Webpage(s) | http://www.research.microsoft.com/~akapoor/papers/NIPS_WS%202010.pdf |
| Alternate Webpage(s) | https://www.microsoft.com/en-us/research/wp-content/uploads/2016/12/NIPS_WS-2010.pdf |
| Alternate Webpage(s) | http://research.microsoft.com/en-us/um/people/akapoor/papers/nips_ws%202010.pdf |
| Alternate Webpage(s) | http://research.microsoft.com/~akapoor/papers/NIPS_WS%202010.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
