Ihaka Lecture Series | Two | Deep learning - why is it deep, and what is it learning? Event as iCalendar

20 March 2019


Venue: Lecture Theatre PLT1, Ground Floor, Building 303, 38 Princes Street, City Campus

Location: Please join us for refreshments from 6pm outside the lecture theatre

Host: Department of Statistics

Cost: Free - all welcome

This lecture replaces the one planned, 'Machine Learning with TensorFlow and R'. 
Our guest speaker, JJ Allaire, is no longer able to present as part of the 2019 Ihaka Lecture Series due to ill-health.

Professor Thomas Lumley, Department of Statistics, Ihaka Lecture Series 2019 speaker
Professor Thomas Lumley

Even ten years ago, neural networks did not have particularly impressive performance as classifiers.

Statisticians regarded them as just one of many black-box approaches to prediction: a relatively unattractive one because of their computational requirements and their opacity. 

Something changed: deep learning is not only trendy, but genuinely superior to older approaches for image classification and generation, and for some other problems. 

Professor Thomas Lumley will talk about how deep convolutional nets are structured and give some intuition for how they can be effective, but also why they are brittle and can fail in remarkably alien ways.

Watch the lecture.


About the speaker

Thomas Lumley is Professor of Biostatistics at the University of Auckland and a member of the R Core Team. He is a user and teacher of machine learning, rather than a researcher, and has an interest in the public understanding and social impact of statistics.

Read more about the Ihaka Lecture Series.

The New Zealand Statistical Association

Our thanks to The New Zealand Statistical Association who are our official sponsors for the 2019 Ihaka Lecture Series.