Symposium Computer Science - Various Speakers

Speaker: Prof. Michel Verleysen, Université catholique de Louvain

Time: 15:00 - 15:30

Title: Nonlinear dimensionality reduction - visualization with machine learning

Abstract:

Nonlinear dimensionality reduction (NLDR) is a branch of the wide machine learning field. NLDR is essentially unsupervised, which means that it is used to find something (information) in data, but we do not know in advance what kind of information. Consequently, even if it is easy to agree on the general principle of reducing the dimension of data without losing too much information, it is very difficult to agree on a scientific measure of how this loss is evaluated, leading to hundreds of NLDR methods. None of them can be objectively considered as better than others; they all reflect a specific user's point of view. Popular methods are often those that come with an efficient implementation, rather than being chosen for the quality (seen by the user) of their outputs.

This talk insists on one of the users' point of views, often underestimated in the literature: the compromise between a global and a local mapping of the data. We show that multiscale methods may produce much more interesting representations, with an additional computational cost that can be limited with the development of fast yet accurate algorithms.

---

Speaker: Prof. Peter Tino, University of Birmingham

Time: 15:30 - 16:00

Title: A Machine Learning Perspective on Driven Dynamical Systems

Abstract:

This presentation will outline a possible framework for studying parametrised input-driven dynamical systems, such as Echo State Recurrent Neural Networks. Rather than generalising concepts from the theory of autonomous dynamical systems, we will be inspired by the theory of feature spaces and kernel machines from the field of Machine Learning. In particular, I will first present ideas of kernel machines as utilised in Machine Learning and then will apply them to analysing input-driven dynamical systems that represent the driving sequences through their state spaces. This viewpoint will enable us to understand properties of such systems that have been empirically observed but so far not theoretically understood. I will also present open problems that such a framework naturally poses.

Reference: P. Tino: Dynamical Systems as Temporal Feature Spaces. Journal of Machine Learning Research, 21(44), 1-42 (2020).

---

Speaker: Reynier Peletier, University of Groningen

Time: 16:00 - 16:20

Title: The nature of jelly-fish galaxies

Abstract:

Clusters of galaxies are large groups of galaxies, bound by gravity. They are the first structures formed in the Universe. Because of their gravity, they continue attracting matter, mostly dwarf galaxies. When such a galaxy falls into the potential well of the cluster, often the dwarf loses gas, which appears as a long tail following the galaxy. Such a galaxy is called a jelly-fish galaxy. In recent years we have learned a lot about such galaxies. In my talk I will give an overview.

---

Speaker: Abolfazl Taghribi, PhD student University of Groningen

Time: 16:20-16:40

Title: Natural computation techniques for uncovering low-dimensional topological structures in large scale astronomical simulations

Abolfazl Taghribi is having his defence on Tuesday November 15 at 16:15 hour

---

Speaker: Dr. Marco Canducci, University of Birmingham

Time: 16:40 - 17:00

Title: The low-dimensional Universe: studying substructures in astronomical particle data sets with 1-DREAM

Abstract:

Low-dimensional structures (one-, two-dimensional) are ubiquitous in astronomical data sets. Be it in particle simulations or observations, filaments or surfaces are always tracers of perturbations in the equilibrium of the studied system and hold essential information on its history and future evolution. The recovery of such structures is often complicated by the presence of a large amount of background and transverse noise in the observational space. 1-DREAM is a newly developed toolbox for the extraction and analysis of low-dimensional, particle data sets. Its application to simulated and observed data allows for a detailed analysis of the local substructures. Their properties are studied in the context of stellar data from GAIA (observed) and gaseous data from dwarf galaxy simulations and simulations of the Large Scale Structures of the Universe.