Learning with Noisy Graph Labels Competition
Learning with Noisy Graph Labels Competition
Embedded Files
🆕 Submission instructions
🆕 Submission instructions
For all partecipants: The deadline for submitting the paper is March 9th.
✨ Top three teams  can submit a 4-8 pages paper for IJCNN 2025 Proceedings on IEEE Xplore.
đź“Ś Other participants can also submit a 2-to-4 page paper to be considered for a poster presentation at the conference (but will not be included in the proceedings).
The paper has to be written  according to the usual IEEE conference template used by the IJCNN 2025 conference (see https://2025.ijcnn.org/authors/initial-author-instructions).Â
In case you are interested these are the instructions for the submission:
Access https://cmt3.research.microsoft.com/IJCNN2025/Submission/IndexÂ
Log in with your credential or create a new account, then select "Create a new Submission" and choose "Competition Track" as for SUBJECT AREAS select "Learning with Noisy Graph Labels".
Description
Description
Handling noisy data is a persistent issue in machine learning, but it takes on a unique complexity in the context of graph structures. In domains where graph data is prevalent—such as social networks, biological networks, and financial systems—noisy labels can significantly degrade model performance, leading to unreliable predictions. Despite its significance, this problem is not well-explored. This competition will address this gap investigating graph classification under label noise. We believe it could drive major advancements in both research and real-world applications.
Competition Papers: the top teams (up to 3) can submit a paper that will be reviewed and can be considered for publication in the IJCNN 2025 Proceedings on IEEE Xplore. Other participants can also submit a 2-to-4 page paper to be considered as poster presentation at the conference but not to be included in the proceedings. 🤩
Competition Papers: the top teams (up to 3) can submit a paper that will be reviewed and can be considered for publication in the IJCNN 2025 Proceedings on IEEE Xplore. Other participants can also submit a 2-to-4 page paper to be considered as poster presentation at the conference but not to be included in the proceedings. 🤩
Submission Instructions
Submission Instructions
The submission consists of two parts :
1) A valid Pull Request on the challenge GitHub page.
2) Submit their solutions to the Hugging Face competition space that hosts the challenge.
The name of the submission must be the same in both the platforms.Â
For the complete set of Rule, please refer to page Rules
Important Dates
Important Dates
Competition openÂ
Competition closes
Winners notificationÂ
Winners papers submission deadline
December 23, 2024
February 20, 2025 12:00:00 AoE
February 22, 2025
March 1, 2025
Evaluation Criteria:Â
The performance of each model will be assessed based on a set of criteria. The evaluation will be conducted using a hierarchical approach, where the first criterion serves as the primary measure of success. The second criterion will only be applied if two or more models achieve equal results on the first criterion and so on.
F1 score on the test dataset provided without ground truth.
Accuracy on the test dataset provided without ground truth.
F1 score on the test set for evaluation, not provided to participants.
Accuracy on the test set for evaluation, not provided to participants.
Inference time measured as the total time taken to generate predictions for all test graphs on our machines.
Join our Discord Server to learn more and discuss about the challenge: Server !💡
Organisers
Farooq Wani
Sapienza University of Rome
He is a third year Ph.D. student. His work revolves around the development of novel approaches for Neural Network architectures resilient to label noise.Â
Maria Sofia Bucarelli
Sapienza University of RomeÂ
She is a PostDoc. Her work focuses on theory and application of machine learning. Her research interests include generalization, learning from noisy labels, and approximation properties of neural networks.
Giulia Di Teodoro
University of Pisa
She is a PostDoc. Her work currently focuses on theory and application of Recommendation systems. During her PhD she focused on Mixed Integer Linear Programming, Precision Medicine and Explainable AI.
Andrea Giuseppe Di Francesco
Sapienza University of Rome
He is a second year Ph.D. student. His work mainly focuses on the design of inductive biases for Graph Neural Networks. He also collaborates with the Institute of Information Science and Technologies of the CNR at Pisa.Â
Page updated
Google Sites
Report abuse