Position: SFI Stokes Lecturer
Office: Office 2.02
Telephone: +353 (0) 1 708 6146
Fax: +353 (0) 1 708 6456
Demos and software:
2012Kaiser C., Pozdnoukhov A., Enabling Real-time City Sensing with Kernel Stream Oracles and MapReduce, Pervasive and Mobile Computing (in print), 2012
M. Batty, K. W. Axhausen, F. Giannotti, A. Pozdnoukhov, A. Bazzani, M. Wachowicz, G. Ouzounis, Y. Portugali. Smart Cities of the Future. The European Physical Journal Special Topics, Volume 214, Issue 1, pp 481-518. 2012
Kling F., Pozdnoukhov A., When a City Tells a Story: Urban Topic Analysis, To appear in ACM proceedings of the 20th ACM SIGSPATIAL GIS, 2012 (Best Poster Award runner up).
Tuia, D., Pozdnoukhov, A., Foresti, L. and Kanevski, M. Active Learning for Monitoring Network Optimization, In Spatio-Temporal Design: Advances in Efficient Data Acquisition (eds J. Mateu and W. G. Mueller), John Wiley & Sons. 2012
McArdle G., Furey E., Lawlor A., Pozdnoukhov A., City-scale Traffic Simulation From Digital Footprints, UrbComp'12 at ACM SIGKDD, 2012
Coffey C., Nair R., Pinelli F., Pozdnoukhov A., Calabrese F. Missed Connections: Quantifying and Optimizing Multimodal Interconnectivity in Cities. To appear in IWCTS of the 20th ACM SIGSPATIAL GIS, 2012
McGrath R., Coffey C., Pozdnoukhov A., Habitualisation: localisation without location data, Nokia MDC challenge at PERVASIVE'2012, 2012
Lawlor A., Coffey C., McGrath R., Pozdnoukhov A., Stratification structure of urban habitats, Pervasive Urban Apps at PERVASIVE'2012, 2012
Foresti L., Kanevski M., Pozdnoukhov A. Kernel-based Mapping of Orographic Rainfall Enhancement in the Swiss Alps as Detected by Weather Radar. IEEE Transactions on Geoscience and Remote Sensing, Issue 99, pp 1-14 2012.
Farmer C., Pozdnoukhov A., Building streaming GIScience from context, theory, and intelligence. Position paper, Big Data Age workshop at GIScience'2012 (our manifesto to GIScience), 2012
2011Tuia D., Joost S., Pozdnoukhov A. Active multiple kernel learning of wind power resources, Machine Learning for Sustainability at NIPS'11, 2011
Pozdnoukhov A., Kaiser C. Scalable Local Regression for Spatial Analytics, Proc of the 19th ACM SIGSPATIAL GIS'2011, 2011 Long paper:
Pozdnoukhov A., Kaiser C. Space-Time Dynamics of Topics in Streaming Text, LBSN at 19th ACM SIGSPATIAL GIS'2011, 2011 (Best Paper Award).
Pozdnoukhov A., Kaiser C. Area-to-point Kernel Regression on Streaming Data, Geostreaming at 19th ACM SIGSPATIAL GIS'2011, 2011
Coffey C., Pozdnoukhov A., Calabrese F. Time of Arrival Predictability Horizons for Public Bus Routes, Computational Transportation Science workshop at 19th ACM SIGSPATIAL GIS'2011, 2011
Walsh F., Pozdnoukhov A., Spatial structure and dynamics of urban communities, Pervasive Urban Applications at PERVASIVE'2011, 2011
Pozdnoukhov, A., Matasci, G., Kanevski, M., and Purves, R.S. Spatio-temporal avalanche forecasting with Support Vector Machines. Nat. Hazards Earth Syst. Sci., 11, 367-382, 2011.
Foresti L., Pozdnoukhov A. Exploration of alpine orographic precipitation patterns with radar image processing and clustering techniques. Meteorological Applications, John Wiley & Sons, DOI 10.1002/met.272, 2011.
Foresti L., Tuia D., Kanevski M. and Pozdnoukhov A. Learning wind fields with multiple kernels. Stochastic Environmental Research and Risk Assessment, Volume 25, Number 1, pp. 51-66, 2011
2010Pozdnoukhov A., Spatial extensions to kernel methods, Proc. of the 18th ACM SIGSPATIAL GIS (short paper), 2010
Pozdnoukhov A., Walsh F., Exploratory Novelty Identification in Human Activity Data Streams, ACM SIGSPATIAL International Workshop on GeoStreaming at 18th ACM SIGSPATIAL GIS, 2010
Pozdnoukhov A., Walsh F., Kaiser F., Statistical Machine Learning from VGI, Position paper at Role of Volunteered Geographic Information in Advancing Science Workshop at GIScience'10, 2010.
Kaiser C., Walsh F., Farmer C. and Pozdnoukhov A., User-centric time-distance representation of road networks. In Springer LNCS proc. of the GIScience'10 (full paper). 2010
Tuia D., Ratle F., Pozdnoukhov A., Camps-Valls G. Multisource Composite Kernels for Urban-Image Classification. IEEE Geoscience and Remote Sensing Letters, Volume 7, Number 1, pp. 88-92, 2010.
Pozdnoukhov A., Dynamic network data exploration through semi-supervised functional embedding. In 17th ACM SIGSPATIAL GIS, 2009
|SFI Research Frontiers Programme: Learning Human Spatial Dynamics. Principle Investigator, (2011-2015).|
|SFI StratAG: Strategic Research Cluster in Advanced Geotechnologies. Co-PI, Scalable Statistical Learning project lead, Coordinator of the City-Scale Demonstrator (2011-2013).|
|GMorphs: Contextual morphing of GMaps (Google Research Award 2010). Principle Investigator.|
| Supervisor, an IBM PhD Fellowship Award to Cathal Coffey.
Data Analytics for Smarter Driving (Scalable Data Analytics Award 2010). Co-PI with PI Tim McCarthy.
|COSMIC: Complexity in Spatial Dynamics (ERA-NET on Complexity). Co-investigator with CASA-UCL (M. Batty), VU Amsterdam (P. Nijkamp), and University of St.Andrews (S. Fotheringham) (2010-2012).|
|Marie-Curie ITN Geocrowd: Creating Geospatial Knowledge World. Scientist in charge at NUIM (2011-2014).|
A micro-simulation of road traffic in Greater Dublin region for a typical weekday. The model is calibrated from census travel surveys as well as social media data and takes into account the geography of community structure of the city. It is described in detail in an upcoming paper .
Dynamic mapping of GSM and 3G mobile phones usage reveal interesting patterns of user's activity. This heatmap is computed with a geographically weighted kernel density estimate with a temporal resolution of 15 minutes. So enjoy one week of Irish life as seen by a mobile phone network!
There are also methods both in spatial statistics and machine learning to downscale population densities to the street level from areal support data which is common to non-pervasive sensing infrastructures .
We use i2maps for a variety of geospatial data analysis studies, including statistical weather now-casting, dynamic mapping of GSM and 3G mobile phones usage, multi-agent simulations of pedestrians and traffic, visualisations of congestion impact on the transportation systems, world-wide instant messaging activity (demo on the left), etc.
Environmental data mining
With Loris Foresti we work on pattern recognition methods in meteorological applications.
Using data streams from weather sensor networks we detect, try to explain and predict such specific events as temperature inversion,
topography influence on mountain winds, and orographic enhancement of precipitation cells detected from rain radar imagery.
|Kanevski M., Pozdnoukhov A., Timonin V. Machine Learning Algorithms for Geospatial Data. Theory, Applications and Software. 377pp. EPFL press, 2009. Link|
Other publications:Demyanov V., Pozdnoukhov A., Christie M., Kanevski M., Detection of Optimal Models in High Dimensional Parameter Space with Support Vector Machines. Springer Series: Quantitative Geology and Geostatistics; ISBN: 978-90-481-2321-6 2010. Link
Pozdnoukhov A., Foresti L. and Kanevski M., Data-driven topo-climatic mapping with machine learning algorithms. Natural Hazards Journal, Volume 50, Issue 3, pp. 497-518. 2009. Link; PDF
Pozdnoukhov A., Purves R.S., Kanevski M. Applying Machine Learning Methods to Avalanche Forecasting. Annals of Glaciology, vol. 49., pp. 107-113, PDF 2008.
Pozdnoukhov A., Kanevski M. Multi-Scale Support vector Regression for hot spot detection and modeling. In Stochastic Environmental Research and Risk Assessment (SERRA), DOI 10.1007/s00477-007-0162-x, 14 pp., 2007.
Pozdnoukhov A., Bengio S. From Samples to Objects: Invariances in Kernel Methods. Pattern Recognition Letters Journal, Volume 27, Issue 10, pp. 1087-1097. 2006.
Pozdnoukhov A., Kanevski M. Monitoring Network Optimisation for Spatial Data Classification Using Support Vector Machines. Int. Journal of Environment and Pollution. Vol.28. 20 pp., 2006.
NCG 602 - Methods and techniques of geocomputation: spatial statistics, geostatistics, GWR.
Project example: ArcGIS toolbox with GUI implementing a custom spatial statistics method.
Scripting languages required: Python, R.
NCG 603 - Advanced methods and techniques: GWR, pattern recognition and machine learning, multi-agent systems.
Project example: Object recognition in optical remote sensing imagery.
Scripting languages required: Python, Matlab.
MS Thesis: I'm always open to supervise a strong thesis. Tell me your idea, check this, be ready to learn a lot and to do some scripting - and expect writing a paper on your reseach afterwards!
Alexei received a Ph.D. in computer science from EPFL, Switzerland, following his research in machine learning methods and computer vision that he carried out at IDIAP Research Institute in Martigny, Switzerland. He then worked at the Institute of Geomatics and Analysis of Risk (IGAR), University of Lausanne.