Differentiable Ranking Metric Using Relaxed Sorting for Top-K Recommendation

Lee, Hyunsung; Cho, Sangwoo; Jang, Yeongjae; Kim, Jae Kwang; Woo, Honguk

doi:10.1109/access.2021.3105389

Cited by 9 publications

(4 citation statements)

References 25 publications

(45 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In terms of applications, differentiable sorting has been leveraged in various contexts, including recommender systems [33], image patch selection [15], selection experts in multi-task learning [25], and attention mechanisms [59]. To the best of our knowledge, the proposed method is the first work to leverage ordering supervision for self-supervised learning of visual representations.…”

Section: Differentiable Sorting and Rankingmentioning

confidence: 99%

Learning by Sorting: Self-supervised Learning with Group Ordering Constraints

Shvetsova¹,

Petersen²,

Kukleva³

et al. 2023

Preprint

View full text Add to dashboard Cite

Contrastive learning has become a prominent ingredient in learning representations from unlabeled data. However, existing methods primarily consider pairwise relations. This paper proposes a new approach towards selfsupervised contrastive learning based on Group Ordering Constraints (GroCo). The GroCo loss leverages the idea of comparing groups of positive and negative images instead of pairs of images. Building on the recent success of differentiable sorting algorithms, group ordering constraints enforce that the distances of all positive samples (a positive group) are smaller than the distances of all negative images (a negative group); thus, enforcing positive samples to gather around an anchor. This leads to a more holistic optimization of the local neighborhoods. We evaluate the proposed setting on a suite of competitive self-supervised learning benchmarks and show that our method is not only competitive to current methods in the case of linear probing but also leads to higher consistency in local representations, as can be seen from a significantly improved k-NN performance across all benchmarks.

show abstract

Section: Differentiable Sorting and Rankingmentioning

confidence: 99%

Learning by Sorting: Self-supervised Learning with Group Ordering Constraints

Shvetsova¹,

Petersen²,

Kukleva³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Applications and Broader Impact. In the domain of recommender systems, Lee et al [11] propose differentiable ranking metrics, and Swezey et al [12] propose PiRank, a learning-to-rank method using differentiable sorting. Other works explore differentiable sorting-based top-k for applications such as differentiable image patch selection [13], differentiable k-nearest-neighbor [1], [14], top-k attention for machine translation [14], and differentiable beam search methods [14], [15].…”

Section: Related Workmentioning

confidence: 99%

Monotonic Differentiable Sorting Networks

Petersen¹,

Borgelt²,

Kuehne³

et al. 2022

Preprint

View full text Add to dashboard Cite

Differentiable sorting algorithms allow training with sorting and ranking supervision, where only the ordering or ranking of samples is known. Various methods have been proposed to address this challenge, ranging from optimal transport-based differentiable Sinkhorn sorting algorithms to making classic sorting networks differentiable. One problem of current differentiable sorting methods is that they are non-monotonic. To address this issue, we propose a novel relaxation of conditional swap operations that guarantees monotonicity in differentiable sorting networks. We introduce a family of sigmoid functions and prove that they produce differentiable sorting networks that are monotonic. Monotonicity ensures that the gradients always have the correct sign, which is an advantage in gradient-based optimization. We demonstrate that monotonic differentiable sorting networks improve upon previous differentiable sorting methods.

show abstract

“…The collaborative filtering recommendation algorithm is the most widely used recommendation algorithm. Its basic idea is to compare the similarities between different users or locations, and then sort the locations based on the similarities, so as to make Top-K recommendations [5]. The content based recommendation algorithm is to analyze the labels of historical data, and then recommending locations of interest to users based on label information.…”

Section: Introductionmentioning

confidence: 99%

Exploring Diversity and Time-aware Recommendations: A LSTM-DNN Model with Bidirectional DTW Algorithm

Li,

Chen,

Sun

et al. 2023

Preprint

View full text Add to dashboard Cite

With the advent of the Web 3.0 era, the number and types of data in the network have sharply increased, and the application scenarios of recommendation algorithms have also been expanded to a certain extent. Location recommendation has gradually become one of the popular application scenarios in recommendation algorithms. Traditional recommendation algorithms not only ignore the time attribute of data when recommending information to users, but also blindly pursue the recommendation accuracy, which will cause certain "information cocoon room" problems. Therefore, this article treats user historical data as a time series and proposes a LSTM-DNN model based on the bidirectional DTW algorithm. Firstly, in response to the issue of different users consuming different amounts of information, this article proposes a bidirectional DTW algorithm to calculate the similarities between different users. Secondly, this article supplements the user dataset from three perspectives: "utilization" of information, "exploration", and spatiotemporal attributes of data, which alleviates the problem of data sparsity and cold start in the dataset to a certain extent. Moreover, it effectively enhances the diversity of recommendation results. Finally, this paper constructs a LSTM-DNN neural network to dynamically obtain user interests and preferences, and proposes a new metric CSSD to measure the diversity of algorithm recommendation results. Experiments have shown that the model effectively enhances the diversity of recommendation results while ensuring recommendation accuracy.

show abstract

Differentiable Ranking Metric Using Relaxed Sorting for Top-K Recommendation

Cited by 9 publications

References 25 publications

Learning by Sorting: Self-supervised Learning with Group Ordering Constraints

Learning by Sorting: Self-supervised Learning with Group Ordering Constraints

Monotonic Differentiable Sorting Networks

Exploring Diversity and Time-aware Recommendations: A LSTM-DNN Model with Bidirectional DTW Algorithm

Contact Info

Product

Resources

About