Preference inference based on maximising margin
| dc.check.embargoformat | Embargo not applicable (If you have not submitted an e-thesis or do not want to request an embargo) | en |
| dc.check.info | Not applicable | en |
| dc.check.opt-out | Not applicable | en |
| dc.check.reason | Not applicable | en |
| dc.check.type | No Embargo Required | |
| dc.contributor.advisor | Wilson, Nic | en |
| dc.contributor.advisor | O'Sullivan, Barry | en |
| dc.contributor.author | Montazery Hedeshi, Mojtaba | |
| dc.contributor.funder | Science Foundation Ireland | en |
| dc.date.accessioned | 2018-08-03T09:17:28Z | |
| dc.date.available | 2018-08-03T09:17:28Z | |
| dc.date.issued | 2018 | |
| dc.date.submitted | 2018 | |
| dc.description.abstract | In a decision-making problem, where we need to choose a particular decision from a set of possible choices, the user often has some preferences which determine if one decision is preferred over another. When the number of choices is large, an intelligent system can help the user by attempting to learn user preferences. One way of learning user preferences is based on the maximum margin approach, where maximising the margin can be seen as satisfying each existing preference input to the greatest degree. In this thesis, we first apply this method to a real-world application, ride-sharing, and examine its potential effectiveness. Nevertheless, we show that the maximum margin preference learning approach is sensitive to the way that preferences inputs and features are scaled. We explain why it is naturally expected that a preference relation is scaling invariant, and go on to construct and characterise some preference relations that are invariant to the scaling of (i) preferences inputs, (ii) features, and (iii) both preferences inputs and features simultaneously. We compare these relations and propose two algorithms to find the optimal elements according to each relation. In the last main chapter, we argue that the rescaling of features is also an issue in the standard SVM classification and propose a new form of more conservative classification that is invariant to the rescaling of features. We argue that this cautious way of classification could be helpful in some critical decision-making applications. | en |
| dc.description.status | Not peer reviewed | en |
| dc.description.version | Accepted Version | |
| dc.format.mimetype | application/pdf | en |
| dc.identifier.citation | Montazery Hedeshi, M. 2018. Preference inference based on maximising margin. PhD Thesis, University College Cork. | en |
| dc.identifier.endpage | 175 | en |
| dc.identifier.uri | https://hdl.handle.net/10468/6572 | |
| dc.language.iso | en | en |
| dc.publisher | University College Cork | en |
| dc.relation.project | info:eu-repo/grantAgreement/SFI/SFI Research Centres/12/RC/2289/IE/INSIGHT - Irelands Big Data and Analytics Research Centre/ | en |
| dc.rights | © 2018, Mojtaba Montazery Hedeshi. | en |
| dc.rights.uri | http://creativecommons.org/licenses/by-nc-nd/3.0/ | en |
| dc.subject | Preference learning | en |
| dc.subject | Preference reasoning | en |
| dc.subject | Artificial intelligence | en |
| dc.thesis.opt-out | false | |
| dc.title | Preference inference based on maximising margin | en |
| dc.type | Doctoral thesis | en |
| dc.type.qualificationlevel | Doctoral | en |
| dc.type.qualificationname | PhD | en |
| ucc.workflow.supervisor | n.wilson@ucc.ie |
Files
Original bundle
1 - 2 of 2
Loading...
- Name:
- Mojtaba Montazery Thesis.pdf
- Size:
- 1.39 MB
- Format:
- Adobe Portable Document Format
- Description:
- Full Text E-thesis
Loading...
- Name:
- Abstract.pdf
- Size:
- 89.58 KB
- Format:
- Adobe Portable Document Format
- Description:
- Abstract
License bundle
1 - 1 of 1
Loading...
- Name:
- license.txt
- Size:
- 5.62 KB
- Format:
- Item-specific license agreed upon to submission
- Description: