Pairwise versus Pointwise Ranking: A Case Study

Melnikov, Vitalik; Hüllermeier, Eyke; Kaimann, Daniel; Frick, Bernd; Gupta, Pritha

doi:10.4467/20838476si.16.006.6187

Cited by 7 publications

(5 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is justified by taking an expectation over all (complete) rankings of X and assuming a uniform distribution. In spite of this apparently oversimplified assumption, and the questionable transformation of ordinal ranks into numerical scores, ERR has shown quite competitive performance in empirical studies, especially when all rankings in the training data (3) are of approximately the same length (Melnikov et al 2016). Given a ranking (1) as training information, the pairwise approach extracts all pairwise preferences x π −1 (i)…”

Section: Methodsmentioning

confidence: 99%

Learning to Rank Based on Analogical Reasoning

Fahandar

Hüllermeier

2018

AAAI

Self Cite

View full text Add to dashboard Cite

Object ranking or "learning to rank" is an important problem in the realm of preference learning. On the basis of training data in the form of a set of rankings of objects represented as feature vectors, the goal is to learn a ranking function that predicts a linear order of any new set of objects. In this paper, we propose a new approach to object ranking based on principles of analogical reasoning. More specifically, our inference pattern is formalized in terms of so-called analogical proportions and can be summarized as follows: Given objects A,B,C,D, if object A is known to be preferred to B, and C relates to D as A relates to B, then C is (supposedly) preferred to D. Our method applies this pattern as a main building block and combines it with ideas and techniques from instance-based learning and rank aggregation. Based on first experimental results for data sets from various domains (sports, education, tourism, etc.), we conclude that our approach is highly competitive. It appears to be specifically interesting in situations in which the objects are coming from different subdomains, and which hence require a kind of knowledge transfer.

show abstract

Section: Methodsmentioning

confidence: 99%

Learning to Rank Based on Analogical Reasoning

Fahandar

Hüllermeier

2018

AAAI

Self Cite

View full text Add to dashboard Cite

show abstract

“…Some PL techniques derive a ranking score, which can be used for pointwise predictions. PL by pairwise comparison has proven to be more robust, providing more stable predictions when the distribution of the labels is not normal or unknown [34], [35].…”

Section: A Preference Learningmentioning

confidence: 99%

Towards General Models of Player Experience: A Study Within Genres

Melhárt¹,

Liapis²,

Yannakakis³

2021

Preprint

View full text Add to dashboard Cite

To which degree can abstract gameplay metrics capture the player experience in a general fashion within a game genre? In this comprehensive study we address this question across three different videogame genres: racing, shooter, and platformer games. Using high-level gameplay features that feed preference learning models we are able to predict arousal accurately across different games of the same genre in a largescale dataset of over 1, 000 arousal-annotated play sessions. Our genre models predict changes in arousal with up to 74% accuracy on average across all genres and 86% in the best cases. We also examine the feature importance during the modelling process and find that time-related features largely contribute to the performance of both game and genre models. The prominence of these game-agnostic features show the importance of the temporal dynamics of the play experience in modelling, but also highlight some of the challenges for the future of general affect modelling in games and beyond.

show abstract

“…Since the pointwise approach optimizes each sample individually, it is flexible on sampling and weighting on instance-level. However, as these scores depend on the observation context, fitting the sample into the fixed score is hard to reflect the inherent ranking property [22].…”

Section: Pointwise Lossmentioning

confidence: 99%

“…This paper tries to combine and complement the two mainstream loss paradigms for the recommendation task. As the pointwise and the pairwise approaches have their advantages and limitations [22], several methods have been proposed to improve the loss function [9,33]. Bellogin et al [1] improved the recommendations based on the formation of user neighbourhoods.…”

Section: Related Workmentioning

confidence: 99%

Learning Explicit User Interest Boundary for Recommendation

Zhuo,

Zhu,

Yue

et al. 2021

Preprint

View full text Add to dashboard Cite

The core objective of modelling recommender systems from implicit feedback is to maximize the positive sample score 𝑠 𝑝 and minimize the negative sample score 𝑠 𝑛 , which can usually be summarized into two paradigms: the pointwise and the pairwise. The pointwise approaches fit each sample with its label individually, which is flexible in weighting and sampling on instance-level but ignores the inherent ranking property. By qualitatively minimizing the relative score 𝑠 𝑛 − 𝑠 𝑝 , the pairwise approaches capture the ranking of samples naturally but suffer from training efficiency. Additionally, both approaches are hard to explicitly provide a personalized decision boundary to determine if users are interested in items unseen. To address those issues, we innovatively introduce an auxiliary score 𝑏 𝑢 for each user to represent the User Interest Boundary(UIB) and individually penalize samples that cross the boundary with pairwise paradigms, i.e., the positive samples whose score is lower than 𝑏 𝑢 and the negative samples whose score is higher than 𝑏 𝑢 . In this way, our approach successfully achieves a hybrid loss of the pointwise and the pairwise to combine the advantages of both. Analytically, we show that our approach can provide a personalized decision boundary and significantly improve the training efficiency without any special sampling strategy. Extensive results show that our approach achieves significant improvements on not only the classical pointwise or pairwise models but also state-of-the-art models with complex loss function and complicated feature encoding. CCS CONCEPTS• Computing methodologies → Ranking; • Information systems → Learning to rank.

show abstract

Pairwise versus Pointwise Ranking: A Case Study

Cited by 7 publications

References 10 publications

Learning to Rank Based on Analogical Reasoning

Learning to Rank Based on Analogical Reasoning

Towards General Models of Player Experience: A Study Within Genres

Learning Explicit User Interest Boundary for Recommendation

Contact Info

Product

Resources

About