We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Information Retrieval

Title: Multi-Sample based Contrastive Loss for Top-k Recommendation

Abstract: The top-k recommendation is a fundamental task in recommendation systems which is generally learned by comparing positive and negative pairs. The Contrastive Loss (CL) is the key in contrastive learning that has received more attention recently and we find it is well suited for top-k recommendations. However, it is a problem that CL treats the importance of the positive and negative samples as the same. On the one hand, CL faces the imbalance problem of one positive sample and many negative samples. On the other hand, positive items are so few in sparser datasets that their importance should be emphasized. Moreover, the other important issue is that the sparse positive items are still not sufficiently utilized in recommendations. So we propose a new data augmentation method by using multiple positive items (or samples) simultaneously with the CL loss function. Therefore, we propose a Multi-Sample based Contrastive Loss (MSCL) function which solves the two problems by balancing the importance of positive and negative samples and data augmentation. And based on the graph convolution network (GCN) method, experimental results demonstrate the state-of-the-art performance of MSCL. The proposed MSCL is simple and can be applied in many methods. We will release our code on GitHub upon the acceptance.
Comments: 12 pages,7 figures,6 tables
Subjects: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
Cite as: arXiv:2109.00217 [cs.IR]
  (or arXiv:2109.00217v1 [cs.IR] for this version)

Submission history

From: Hao Tang [view email]
[v1] Wed, 1 Sep 2021 07:32:13 GMT (1483kb,D)

Link back to: arXiv, form interface, contact.