Invited talk, Continual AI Meetup, Virtual, Everywhere
Talks and presentations
Keynote, Private NLP Workshop, WSDM 2020, Houston, Texas
Invited talk, Interactive AI CDT Winter School, University of Bristol, Bristol, UK
Invited talk, NeurIPS meetup, MathWorks, Cambridge, UK
In the real world, we are often faced with situations where data distributions are shifting over time, and we would like to make predictions about things we have never seen before. These situations fall under the umbrella of “Continual Learning”, which has gained popularity in recent years in the AI/ML community. Researchers have focused methods that can adapt to new tasks while preventing “catastrophic forgetting” (e.g. in neural networks) through regularization, generative replay, approximate Bayesian approaches etc. I argue, however, that these are far away from the reality of production environments. In this talk I will present efforts we have been making to bridge this gap, tackling the problem from both research and engineering perspectives.
Invited talk, Colloquium of the Institute of Cognitive Science, Institute of Cognitive Science, Osnabrück, Germany
How do we have machines that learn continually, as data arrives? Continual (aka lifelong) learning has gained popularity in recent years in the ML community, in particular methods that try to prevent “catastrophic forgetting” in neural networks. Whilst there are compelling advances in the field, I argue that these are far away from the reality of production environments. In this talk I will present efforts we have been making to bridge this gap, tackling the problem from both research and engineering perspectives.
Invited talk, Bristol Robotics Lab, Bristol, UK
Invited talk, South West Data Meetup, Pervasive Media Studio, The Watershed, Bristol
Invited Talk, 2nd Workshop on Applications of Pattern Analysis (WAPA), Castro Urdiales, Spain
Invited Talk, 6th International Summer School on Pattern Recognition, Plymouth, UK
The course is intended to give an overview of the kernel approach to pattern analysis. This will cover:
- Why linear pattern functions?
- Why kernel approach?
- How to plug and play with the different components of a kernel-based pattern analysis system?