Universiteit Leiden

nl en

Lecture

This Week’s Discoveries | 19 March 2019

Date
19 March 2019
Time
Series
This Week's Discoveries
Location
Oort
Niels Bohrweg 2
2333 CA Leiden
Room
De Sitterzaal

First lecture

Title
Automatic Configuration for Improving Grammatical Evolution


Speaker
Hao Wang  (LIACS)
Hao Wang is a postdoctoral researcher and member of the Natural Computing group at LIACS. He received his master’s degree in Computer Science from Leiden University in 2013 and obtained his PhD (cum laude, promotor: Prof. Thomas Ba
̈ck) in Computer Science from the same university in 2018. His research interests are proposing, improving and analyzing stochastic optimization algorithms, especially Evolutionary Strategies and Bayesian Optimization. In addition, he also works on developing statistical machine learning algorithms for big and complex industrial data. He also aims at combining the state-of-the-art optimization algorithm with data mining/machine learning techniques to make real-world optimization tasks more efficient and robust.

Abstract
Grammatical Evolution (GE) systems have been invented to evolve computer programs, aiming at solving real-world problems. Such a system has a number of hyper-parameters that control its behaviour, which have a significant impact on the performance of the GE system. Moreover, the optimal configuration of those parameter varies from one application to another and not much human experience is available to deal with this issue. In this talk, I will present a recent approach - Automatic Configuration of hyper-parameters of Grammatical Evolution systems - which is built on one of our optimization algorithms, called Mixed-Integer Parallel Efficient Global Optimization. This approach is validated on four common test problems used in the Grammatical Evolution community: StringMatch, symbolic regression (the ‘Vladislavleva-4’ problem), banknote classification and the so-called Pymax task. The experimental results show that the average performance of the GE system is improved significantly (between 25% and 168%) on all of the test problems. In addition, the resulting overall best hyper-parameter settings are substantially different from the commonly used ones.

Second lecture

Title
Lighting-Up Dark Matter haloes with GAEA

Speaker
Gabriella DeLucia (OATs)
Gabriella is a Senior Researcher of Theoretical Astrophysics at INAF - Astronomical Observatory of Trieste (OATs). She is one of the organizers of the workshop  “Galaxy Evolution in the Cosmic Web”
that is being held in the Lorentz Center from 18 March through 22 March.

Abstract
I will give a very brief overview of semi-analytic methods adopted to model galaxy formation and evolution in a cosmological framework. I will then focus on the latest results from our Galaxy Evolution and Assembly model, highlighting successes and open issues.

Read more about the lecture series This Week's Discoveries

This website uses cookies. More information