Nikolaos (Nikos) Nikolaou
Selected Presentations
The list below is outdated. For paper presentations, go to the
publications page.
2018 - 2019
Correcting Transiting Exoplanet Light
Curves for Stellar Spots: The 1st ARIEL Machine Learning Competition
A description of the machine learning challenge I organised in the context of the upcoming
ESA M4 Ariel Space Mission.
The challenge was hosted by the European Conference on Machine Learning (ECML-PKDD 2019).
The talk discusses the problem, the motivation, the data, organizational aspects and the very promising initial baseline
solutions our team developped. It was presented in the UK Exoplanets Meeting 2019 (UKEXOM 2019), London, UK. An earlier
version of the talk was given in the ARIEL Space Mission Consortium in Palermo (March 2019). Stay posted for the update
in the European Week of Astronomy & Space Science 2019, Lyon, France (June 2019).
AI Techniques for Exoplanets (Part 1
& Part 2)
A lecture for the Exoplanets postgraduate course of the Department of Physics and Astronomy of the
University College London, UK. It is meant as a gentle introduction to Artificial Intelligence and in particular
Machine Learning and its applications in the field of Exoplanetary Science.
Margin Maximization as Lossless Maximal Compression
My latest research on the theoretical side of machine learning, connecting hypothesis margin maximization in
classification to information theory. Slides to be uploaded soon. Presented in the Greek Stochastics 2018 meeting
in Athens, Greece and in the School of Mathematics, Statistics and Actuarial Science of the University of Kent, UK.
Introduction to Machine Learning
Introductory talk, the first of the Machine Learning PAD seminar series of the Department of Physics & Astronomy
of the University College London, UK.
2016 - 2017
Learning from Imbalanced Classes: Problem Statement & Methods
Talk contributed to Workshop on Class Imbalance in Machine Learning Classification, organized by the
4IR STFC Centre for Doctoral Training in Data Intensive Science of the University of Manchester, UK. An
updated version
was delivered in February 2019 as part of the Machine Learning PAD seminar series of the Department of Physics & Astronomy of the University College London, UK.
Boosting for Probability Estimation & Cost-Sensitive Learning
Lecture on the work I carried out during my PhD, along with online extensions and connections to other areas of machine learning I am currently exploring.
This is the full version of the talk (~ 1.5h). Appropriately adapted, shorter versions were delivered in:
- the Astrophysics group of the Department of Physics & Astronomy of the University College London, UK
- the AINC seminar series of the School of Computer Science of the University of Birmingham, UK
- the Statistical Machine Learning group of the Department of Computing of the Imperial College London, UK
- the Department of Informatics and Telecommunications of the University of Athens, Greece
- the Department of Informatics of the Athens University of Economics and Business, Greece
- the Institute for Adaptive and Neural Computation of the Department of Informatics of the University of Edinburgh, UK
- the Department of Computer Science of the University of Cyprus, Cyprus
- the Electrical Engineering, Computer Engineering and Informatics of the Cyprus University of Technology, Cyprus
- the Greek Stochastics 2017, Greece.
2015
Asymmetric boosting algorithms: Do we really need them?
Lecture on my latest research delivered in the Research Symposium of the University of Manchester, UK.
Cost-sensitive learning with AdaBoost
A brief introduction to cost-sensitive learning, followed by my latest research on cost-sensitive AdaBoost, delivered to the postgraduate class of
COMP61011: Foundations of Machine Learning of the University of Manchester, UK.
Optimal inductive inference and its approximations
Two-part talk on inductive inference delivered to the members of the
Machine Learning and Optimization research group of the University of Manchester, UK.
We first gave an intuitive interpretation of Solomonoff Induction, an intractable formalization of optimal inductive inference which combines ideas from philosophy, computer science, statistics & information theory.
We then saw how nature and machine learning overcome the ill-posedness of induction by introducing assumptions and settling for approximations. Concepts covered included: common assumptions, bias-variance
tradeoff, inductive bias & the no-free-lunch theorems, the role of Occam's Razor and elements of statistical learning theory.
2014
Introduction to AdaBoost
Introductory lecture on AdaBoost delivered to the postgraduate class of COMP61011:
Foundations of Machine Learning of the University of Manchester, UK.