Local intrinsic dimensionality and its applications for anomaly detection and self supervised learning
Local intrinsic dimensionality and its applications for anomaly detection and self supervised learning
In this seminar, we will review a measure known as Local Intrinsic Dimensionality (LID), which can be used for characterizing the complexity of local neighbourhoods in data. LID can loosely be thought of as a measure for the number of latent variables needed to describe a particular location in multi dimensional space. In this talk we will review the LID measure and its uses in machine learning and data mining. In particular, we focus on two recent exciting applications.
The first application is in anomaly detection, where we introduce a ‘dimensionality-aware’ outlier detection method, DAO, which is derived as an estimator of an asymptotic local expected density ratio involving a query point and a close neighbor drawn at random. DAO significantly outperforms three popular and important benchmark local outlier detection methods.
The second application is in the field of self supervised learning, where we show i) how the use of LID for dimensionality regularization at a local level can be used to mitigate an underfilling phenomenon known as dimensional collapse and ii) how the local dimensionality of deep representations can be used as a proxy target when searching for suitable data augmentation policies in contrastive learning
Speaker
Professor James Bailey
James Bailey is a Professor in the Faculty of Engineering and Information Technology at The University of Melbourne and Program Lead for the University’s Artificial Intelligence Platform. He was previously an Australian Research Council Future Fellow and is a researcher in the field of machine learning and artificial intelligence, including interdisciplinary applications and operational frameworks. His interests particularly relate to the assurance, certification and safety of systems based on machine learning and artificial intelligence. He works on the deployment of AI systems in collaboration with a wide range of industry and government partners across sectors such as defence, energy and health. He is co-General Chair for the Australasian Joint Conference in Artificial Intelligence to be held in Melbourne in December 2024.
Monash University values the privacy of every individual's personal information and is committed to the protection of that information from unauthorised use and disclosure except where permitted by law. For information about the handling of your personal information please see Data Protection and Privacy Procedure and the relevant Data Protection and Privacy Collection Statement that applies to you depending on the nature of your interaction with us.
If you have any questions about how Monash University is collecting and handling your personal information, please contact our Data Protection and Privacy Office at dataprotectionofficer@monash.edu.