Reproduce “Generalized Top-k Mallows Model for Ranked Choices” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shahrzad Haddadan, Sara Ahmadian
The Generalized Top-k Mallows Model (TopKGMM) is introduced, along with efficient algorithms for sampling, choice probability computation, and active learning of its parameters from partial choice data. This model demonstrates superior predictive accuracy compared to the Multinomial Logit model in real-world choice prediction tasks.
Key insights
- PRIM successfully addresses the open problem of efficient, RIM-like sampling for Top-k Mallows models, providing a computationally viable method for generating samples.
- The TopKGMM, with its product-specific weights, offers a more flexible and realistic representation of partial user preferences than previous Top-k Mallows variants, better reflecting real-world choice behavior.
- The BUCCHOI active learning algorithm provides a theoretically grounded and efficient approach to inferring complex preference parameters (center ranking and k) from limited choice observations, requiring a small number of samples.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.