Files
Winand_Maxime_29301500_2021.pdf
Closed access - Adobe PDF
- 677.87 KB
Winand_Maxime_29301500_2021_APPENDIX1.zip
Closed access - Unknown
- 26.16 KB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- We are interested in inferring the inherent ratings of a set of items based onpairwise comparisons among them. More specifically, we expand on the results ofHendrickx et al.,2020 [9] and handle three problematics about pairwise comparisons.We analyze the impact of the graph resistance on the error rate of the methods. Weempirically and formally derive the error bounds that were theoretically computedpreviously. We show that they don’t hold if either the set of item or the numberof comparisons is too small. We design two tools to select the best performingmethod for a given situation: one based on a preliminary estimation of the weightsand the other on a decision tree. Knowing that the empirical least square methodoutperforms the others, we explain and demonstrate its surprising performances
Citations
Winand, M. (2021). Learning from pairwise comparisons. https://dial-mem.test.bib.ucl.ac.be/handle/2078.2/22247