Skip to Content

PhD-Position ‘Machine Learning and Data Mining Methods for High-Dimensional Spectral Data’


PhD position in the field of chemometric modeling

The basic task is to develop enhanced methods for chemometric calibration models, which are used in the field of process analytic chemistry in order to perform automatic quantification of concentrations of relevant chemical substances. Apart from aiming at accurate quantification statements, other important goals include anomaly detection, prediction of future states and convergence of batch processes.

The data source for setting up calibration models are spectral data and hyper-spectral images (3-D spectral data). The current situation is that basically linear methods (such as Principal Component Analysis, Partial Least Squares or Ridge Regression) are used for estimating calibration models from the data. The task is now to develop enhanced non-linear machine learning models and/or data-driven modeling techniques in order to boost quantification performance. These should be able to deal with high-dimensional feature vectors (each sample in a spectrum corresponds to one dimension) and usually a low number of data samples (the collection of spectra is quite time-intensive).
Once improved performance of (quantitative and predictive) calibration models could be verified based on samples collected from real chemical processes, in a second step, adaptive methods for updating, refining, extending the models on-the-fly during on-line operation mode should be investigated and developed. This is necessary in order to automatically account for the changing dynamics at the chemical processes.

Profile/Personal Qualification of the candidate:
Ideally, the candidate should have a completed master in Mathematics/Statistics/Informatics/Mechatronics or other technical study and already have some knowledge in data-driven modeling, machine learning and/or data mining.

Duration of contract: 4 years , Start Date: 1st of September 2010

Contact: Dr. Edwin Lughofer, email: Tel.: +43 (0)7236 3343 435