Lezing | Seminar
Data Science Onderzoeksprogramma
- 18 mei 2018
- Science Campus
2333 CC Leiden
Data Science seminar bij de Faculteit Wiskunde en Natuurwetenschappen
Vanuit het universiteitsbrede Data Science Onderzoeksprogramma worden enkele malen per jaar seminars georganiseerd, roulerend over de faculteiten. De promovendi van de betreffende faculteit presenteren in interactie met het publiek hun onderzoek en tonen de veelzijdige manier waarop zij Data Science toepassen. Op 18 mei organiseert de Faculteit Wiskunde en Natuurwetenschappen dit evenement. De middag wordt afgesloten met een borrel. Iedereen is van harte welkom!
Registreer via deze link.
Presentatie Nuno César de Sá: Towards an evidence-based Rewilding conservation through Remote Sensing and Data Science
As one of the increasingly accepted approach to conservation, Rewilding aims to restore 'lost' ecosystems by reintroducing extinct/surrogate species into interconnected core nature areas with minimal management and human presence. Rewilding conservationists belief that not only this approach can help to create hotspots of biodiversity protected by a resilient ecosystem, but also that these systems are self-sustainable. Opponents often criticize the lack of evidence-based research to support Rewilding conservation, with some pointing out that up to 70% of the reintroductions fail and that this conservation strategy fails to address the socio-economical and ethical impacts of large-scale 'ecosystem engineering'.
Therefore there is a need for the development of data-driven models that collect and integrate data from both the vegetation and animals into a common framework and allow a more complete understanding of the spatio-temporal dynamics of ecosystems. The PhD research consists of addressing the challenges of integrating multiple Remote Sensing data sources with detection and tracking systems to contribute towards an evidence-based decision making process in Rewilding conservation.
Presentatie Laura Zwep: High-dimensional data of missingness.
Large molecular 'omics' datasets based on samples collected from large patient cohorts are increasingly being produced in order to further elucidate disease mechanisms, and to identify novel biomarkers to guide drug treatment. However, meaningful statistical analysis of such datasets is associated with several important challenges. This project will primarily focus on two particular challenges related to the selection of predictive disease biomarkers across multiple omics data types, and solutions to address missingness of data in such datasets.