Matching Novelty While Training: Novel Recommendation Based on Personalized Pairwise Loss Weighting

Lo, Kachun; Ishigaki, Tsukasa

doi:10.1109/icdm.2019.00057

Cited by 7 publications

(9 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second group of methods adopt end-to-end models. For example, Lo et al [6] propose a personalized pairwise novelty weighting for BPR loss function as an end-to-end method. Liu and Zheng [7] present a network architecture for long-tail session-based recommendation by introducing an adjustable preference mechanism.…”

Section: Novel Recommendationmentioning

confidence: 99%

“…The main target for novel recommendation is to optimize the trade-off between accuracy and novelty [5], [6], [7]. On one hand, the recommendation accuracy is evaluated by a matching score between the recommendation list and the ground-truth list.…”

Section: Problem Formulationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

“…Several recent studies [6], [7], [8] propose the end-toend framework where the recommendation is optimized for both the accuracy and novelty at the same time. While making progress, the PPNW and TailNet methods in [6], [7] tend to recommend tail items and result in a relatively low accuracy. More importantly, none of the re-ranking and the end-to-end based novel recommendation methods differentiate the degree of users' intent to popularity.…”

Section: Introductionmentioning

confidence: 99%

“…Kapoor et al [11] point out in their pioneering work that an item can be novel in three ways: (1) it is new to the system and thus for every user (cold-start), (2) it is known to the system but new to the single user, (3) it is known to the user long before but forgotten at the moment. Unfortunately, existing studies [1], [2], [3], [4], [5], [6], [7], [8] focus on novel recommendation tasks under the above definition (2) (3), and none of them has ever taken cold-start items into consideration. Meanwhile, previous research on cold-start recommendation [12], [13], [14], [15], [16], [17], [18] does not include long-tail items, either.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Intent Disentanglement and Feature Self-supervision for Novel Recommendation

Qian¹,

Liang²,

Li³

et al. 2021

Preprint

View full text Add to dashboard Cite

One key property in recommender systems is the long-tail distribution in user-item interactions where most items only have few user feedback. Improving the recommendation of tail items can promote novelty and bring positive effects to both users and providers, and thus is a desirable property of recommender systems. Current novel recommendation studies over-emphasize the importance of tail items without differentiating the degree of users' intent on popularity and often incur a sharp decline of accuracy. Moreover, none of existing methods has ever taken the extreme case of tail items, i.e., cold-start items without any interaction, into consideration.In this work, we first disclose the mechanism that drives a user's interaction towards popular or niche items by disentangling her intent into conformity influence (popularity) and personal interests (preference). We then present a unified end-to-end framework to simultaneously optimize accuracy and novelty targets based on the disentangled intent of popularity and that of preference. We further develop a new paradigm for novel recommendation of cold-start items which exploits the self-supervised learning technique to model the correlation between collaborative features and content features. We conduct extensive experimental results on three real-world datasets. The results demonstrate that our proposed model yields significant improvements over the state-of-the-art baselines in terms of accuracy, novelty, coverage, and trade-off.

show abstract

Section: Novel Recommendationmentioning

confidence: 99%