 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
Division of Mathematical Methods of Information Technology - Research
One methodology studied is nonparametric function estimation, in particular probability density estimation and regression. Of central interest are the so-called kernel estimation methods. Often the price paid for using such methods in their basic form is heavy computation and one important goal is to reduce the amount of required computation without losing much in flexibility and estimation quality. Neural computing, the mathematical analysis
and application of artificial neural networks, offers another interesting
methodology for the analysis of complex data. In function estimation problems,
one characteristic feature of neural computing methods is their high flexibility
that manifests itself in the rich class of functions neural models are able
to capture. From a mathematical point of view the flexibility of neural networks
therefore often likens them to nonparametric function estimation methods.
Applications Projects Research collaboration is underway in the MODAFOR joint project with the Finnish Forest Research Institute (E. Tomppo) and the Laboratory of Space Technology of the Helsinki University of Technology (M. Hallikainen) on modeling and data-analysis for satellite based forest inventory. Another active project is the joint work with the ECRU unit (A. Korhola) of the Department of Ecology and Systematics of the University of Helsinki on Holocene temperature reconstruction. We also collaborate with the researchers of the National Public Health Institute (J. Haukka) on data-analysis of neuropsychological testing of schizophrenia. In industrial collaboration active work is underway with Instrumentarium Imaging. Index | Research |
Publications | Brief
history | Educational goals | Staff |
  |
