With the rise of large Al models, dataho02eassets have gained increasingimportance. Understanding how toidentify and correct label errors in ourdatasets is crucial. This is because labelerrors are pervasive in the era of bigdata and rectifying them canProf. Tongliang Liusignificantly enhance our knowledgeARC Future FellowMoreover, large Al models are prone toUSYD, Australiaoverfitting label errors, which hinders their ability togeneralize. In this talk, we will present typical approachesto handle label noise, such as extracting confidentexamples and modelling the label noise. By illustrating theintuitions behind state-of-the-art techniques, we wouldequip researchers and practitioners with valuable insightsinto effectively managing label noise.

  Bio: Tongliang Liu is the Director of Sydney Al Centre at The University of Sydney, Australia. He is broadly interested in the fields of trustworthy machine learning and its interdisciplinary applications, with a particular emphasis on learning with noisy labels, adversarial learning, causal representation learning, 3D computer vision, and Al for science. He is/was a senior meta reviewer for many conferences, such