25 January 2019 from 15:00 to 17:00

DSCC Central Topic Seminar #6: Machine Learning Applications in Geoscience and Risk Assessment

The Data Science & Complexity Centre (DSCC) Central Topic Seminars are a series of seminars co-organized by the Utrecht Applied Data Science, the Utrecht Bioinformatics Center, and the Centre for Complex Systems Studies. It will consist of tutorials, excursions, software training and specialist lectures. We aim to expose the central topic "Machine Learning" for the broad community within and outside Utrecht University and bring the researchers from different backgrounds together.

In the first hour, Dr. Jiong Wang from the Department of Physical Geography, will give a specialist lecture on Machine Learning Applications in Geoscience titled "Deep learning for mapping and understanding informal human settlements".

Abstract: Machine learning techniques have been frequently applied in mapping human settlements in very high-resolution satellite images. Among these, deep convolutional neural networks (DCNNs) have shown exceptional efficiency in automatic settlement mapping. Yet DCNNs have never been examined in mapping very small heterogeneous deprivation areas (pockets) at large scale. In this talk, a deep learning architecture is introduced and evaluated in the case of deprived areas in cities. This talk will show the sensitivity of DCNNs to the spatial information contained in remotely sensed images, and how well-designed DCNNs can map urban deprivation at the city scale with limited training data.

In the second hour, Jules Kerckhoffs from the Institute for Risk Assessment Sciences will give a specialist lecture on Machine Learning Applications in Risk Assessment titled "Performance of Prediction Algorithms for Modelling Air Pollution".

Abstract: This talk will be about creating Land use Regression (LUR) models for air pollutants. These are often developed using multiple linear regression techniques. However, in the past decade linear (stepwise) regression methods have been criticized for their lack of flexibility, ignoring potential interaction between predictors and limited ability to incorporate highly correlated predictors. We used ultrafine particles (UFP) data, collected with a mobile platform, to evaluate range of different modelling approaches (such as LASSO, Additive models, Random Forest and Boosting) to estimate long-term UFP concentrations by estimating precision and bias based on an independent external dataset.

Both students and staff are welcome.

DSCC Central Topic Seminars:

Some of the Seminars are available to watch via the CCSS YouTube channel.

Venue: Boothzaal, University Library, De Uithof, Utrecht

Please register before Thursday 24 January 2019.

Start date and time
25 January 2019 15:00
End date and time
25 January 2019 17:00