Uploaded October 2021 | Updated September 2026, 22 hours ago
October 2021, Tom Dietterich, Emeritus Professor of Computer Science at Oregon State University, and considered one of the pioneers in the machine learning field, gave a keynote presentation at Amazon's annual machine learning conference.
Thomas discussed how every deployed learning system should be accompanied by a competence model that can detect when new queries fall outside its region of competence.
His presentation explores the application of anomaly detection to provide a competence model for object classification in deep learning. He considers two threats to competence: queries that are out-of-distribution and queries that correspond to novel classes.
Thomas reviews the four main strategies for anomaly detection and then surveys some of the many recently-published methods for anomaly detection in deep learning. The central challenge is to learn a representation that assigns distinct representations to the anomalies.
The talk concludes with a discussion of how to set the anomaly detection threshold to achieve a desired missed-alarm rate without relying on labeled anomaly data.
Follow us:
Website: https://www.amazon.science
Twitter: twitter.com/AmazonScience
Facebook: facebook.com/AmazonScience
Instagram: instagram.com/AmazonScience
LinkedIn: linkedin.com/showcase/AmazonScience
Newsletter: https://www.amazon.science/newsletter
#AmazonScience #MachineLearning
October 2021, Tom Dietterich, Emeritus Professor of Computer Science at Oregon State University, and considered one of the pioneers in the machine learning field, gave a keynote presentation at Amazon's annual machine learning conference.
Thomas discussed how every deployed learning system should be accompanied by a competence model that can detect when new queries fall outside its region of competence.
His presentation explores the application of anomaly detection to provide a competence model for object classification in deep learning. He considers two threats to competence: queries that are out-of-distribution and queries that correspond to novel classes.
Thomas reviews the four main strategies for anomaly detection and then surveys some of the many recently-published methods for anomaly detection in deep learning. The central challenge is to learn a representation that assigns distinct representations to the anomalies.
The talk concludes with a discussion of how to set the anomaly detection threshold to achieve a desired missed-alarm rate without relying on labeled anomaly data.
Follow us:
Website: https://www.amazon.science
Twitter: twitter.com/AmazonScience
Facebook: facebook.com/AmazonScience
Instagram: instagram.com/AmazonScience
LinkedIn: linkedin.com/showcase/AmazonScience
Newsletter: https://www.amazon.science/newsletter
#AmazonScience #MachineLearning
