Universiteit Leiden

nl en


This Week’s Discoveries | 10 April 2018

10 april 2018
Niels Bohrweg 2
2333 CA Leiden

First Lecture

Sequence-to-sequence learning with neural networks: opening the black box.

Arianna Bisazza (LIACS) Arianna is assistant professor at LIACS since 2017. Her research focuses on the statistical modeling of natural language, with the main goal of improving the quality of machine translation for challenging language pairs. She previously worked as a postdoc at the University of Amsterdam and as a research assistant at Fondazione Bruno Kessler, Italy. She obtained her PhD from the University of Trento in 2013 and was awarded a VENI grant in 2016.

Sequence-to-sequence models based on neural networks have recently led to unprecedented high-quality results in a number of challenging AI tasks like machine translation, text summarization, speech recognition, speech synthesis and generation of video descriptions. But how can we systematically detect and understand the strengths and potential weaknesses of such models? In this talk I will explain how these models work and give an overview of my current research around this crucial question.


Second Lecture

Paving the route towards running the simulations required for the interpretation of future cosmological probes

Matthieu Schaller (Leiden Observatory) Matthieu is an associate researcher in the group of Joop Schaye at the Leiden Observatory. He received a VENI grant is 2017 for “Evaluating the effect of baryons on cosmological probes with next-generation simulations”.

Large international cosmology and dark matter detection experiments that will produce data over the next decade rely heavily on numerical simulation for their interpretation. To achieve the required accuracy and correctly include the back-reaction of the baryons on the dark matter, these simulations need to model volumes too large to be simulated at the required resolution with current numerical codes. The challenge is to harvest all the computing power available on modern super-computer to obtain as detail maps of a virtual cosmos as possible. This talk sits at the cross-point between astrophysics, cosmology and high-performance computing and I will discuss how, in the next few years, we are planning to tackle the challenge outlined above.