# Performance of Hyperbolic Geometry Models on Top-N Recommendation Tasks

**Authors:** Leyla Mirvakhabova, Evgeny Frolov, Valentin Khrulkov, Ivan Oseledets, Alexander Tuzhilin  
**arXiv:** [2008.06716](https://arxiv.org/abs/2008.06716)  
**Date:** 2020-08-15  
**Summary page:** https://www.summarizepaper.com/en/paper/2008.06716v1/

## Key points

- Paper title: "Performance of Hyperbolic Geometry Models on Top-N Recommendation Tasks"
- Authors: Leyla Mirvakhabova, Evgeny Frolov, Valentin Khrulkov, Ivan Oseledets, Alexander Tuzhilin
- Approach: Utilizes hyperbolic geometry for collaborative filtering
- Methodology: Simple autoencoder with a single hidden layer
- Performance: Outperforms traditional Euclidean models and competes with state-of-the-art techniques
- Impact of space curvature on model effectiveness explored
- Optimization of curvature parameter through data-driven methods discussed
- Accepted at ACM RecSys 2020 and spans 7 pages
- Potential of hyperbolic geometry in enhancing recommendation systems highlighted

## Summary

The paper "Performance of Hyperbolic Geometry Models on Top-N Recommendation Tasks" by Leyla Mirvakhabova, Evgeny Frolov, Valentin Khrulkov, Ivan Oseledets, and Alexander Tuzhilin presents a novel approach to collaborative filtering using hyperbolic geometry. Their method utilizes a simple autoencoder with a single hidden layer and outperforms traditional Euclidean models while also competing with state-of-the-art techniques. The study delves into the impact of space curvature on the effectiveness of hyperbolic models and offers insights on optimizing this parameter through data-driven methods. Accepted at ACM RecSys 2020 and spanning 7 pages, this research sheds light on the potential of hyperbolic geometry in enhancing recommendation systems and highlights the efficiency of their proposed methodology in achieving optimal results.