Mixed Dyad Ranking (Ba/Ma)

Topic for a bachelor/master's thesis

Short Description:

The contextual dyad ranking framework is a generalization of the label ranking problem, where the instances and additionally also the alternatives are represented in terms of attributes. In [1] a dyad ranking learner is proposed by maximizing the likelihood-function of a bilinear Plackett-Luce model, which is an extension of a statistical model used for label ranking taking the additional information of the attributes into account. Recently, there have been some works on mixtures of Plackett-Luce models, which have shown to give a better fit to the data [2]. However, a mixture of bilinear Plackett-Luce models was not considered so far.

The goal of the thesis is to introduce a mixture model of bilinear Plackett-Luce models and investigate the identifiability of the model parameters. As a next step, an adaptation of the famous Expectation-Maximization (EM) algorithm, which is a popular framework for fitting the parameters of parametric mixture models, should be investigated.

Requirements

The requirements for the thesis include the following: Implementation of a dyad ranking mixture learner and an experimental study of its performance on synthetic as well as real-word data. In particular, its superiority over classical ranking methods should be demonstrated.

Prerequisites

Expertise in machine learning or statistical learning theory, programming skills.

Contact

Prof. Eyke Hüllermeier or Viktor Bengs

References

• [1] D. Schäfer and E. Hüllermeier. Dyad Ranking using Plackett-Luce Models based on Joint Feature Representations. Machine Learning, 107(5):903–941, 2018.
• [2] A. Liu, Z. Zhao, C. Liao, P. Lu and L. Xia Learning Plackett-Luce mixtures from partial preferences. In Proceedings of AAAI,, vol. 127, 2019.