Improving learning with noisy labels in two possibile scenarios.

Add to your list(s) Download to your calendar using vCal

• Maria Sofia Bucarelli, La Sapienza University
• Tuesday 23 May 2023, 17:30-18:00
• Lecture Theatre 2.

If you have a question about this talk, please contact Pietro Lio.

Learning from data with noisy labels is a challenging problem that arises in various practical applications. Noisy data can indeed arise in various real-world problems, such as medical diagnosis, autonomous driving, fraud detection, and natural language processing. Its presence can significantly impact the accuracy and reliability of machine learning models. In this talk, we will introduce two different frameworks for improving learning with noisy labels in two possible scenarios. In the first scenario, we assume access to data labeled by multiple annotators. In the second scenario, only one label is given for each sample. For the first case, we will leverage inter-rater agreement to effectively mitigate the issue of noisy labels. In the second scenario, our framework proposes a novel approach that combines the use of class centroids and an outlier discounting strategy.

This talk is part of the Artificial Intelligence Research Group Talks (Computer Laboratory) series.

This talk is included in these lists:

• All Talks (aka the CURE list)
• Artificial Intelligence Research Group Talks (Computer Laboratory)
• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge Forum of Science and Humanities
• Cambridge Language Sciences
• Cambridge talks
• Chris Davis' list
• Department of Computer Science and Technology talks and seminars
• Guy Emerson's list
• Hanchen DaDaDash
• Interested Talks
• Lecture Theatre 2
• School of Technology
• Speech Seminars
• Trust & Technology Initiative - interesting events
• bld31
• ndk22's list
• ob366-ai4er
• rp587
• yk373's list

Note that ex-directory lists are not shown.