My Scientific Activity
Site under construction. Some links can be dead.
News:
no news!
List of Positions Held and Scientific Activity in Short
Since 16.03.2020 till now: Independent self-supporting researcher.
Since 5.10.2019 till 15.03.2020 I worked as a lecturer at Wyższa Szkoła Bankowa (WSB) in Toruń, Poland.
Since 1.03.2003 till 28.02.2018 I worked at the Institute of Physics at Kazimierz Wielki University, Bydgoszcz, Poland
Since 2004 till 2008 I had a second position at the University of Economy, Bydgoszcz, Poland
From 1996 till 2002 I had been a Ph.D. student of Physics at the Nicholas Copernicus University, Toruń, Poland. | Prof. Włodzisław Duch was my Ph.D. supervisor.
From 1990 till 1995 I was a student of Physics at the Nicholas Copernicus University, Toruń, Poland. | Prof. Brian G. Wybourne was my M.Sc. supervisor.
Here is my Curriculum Vitae (Polish,English)
Here is my Google Scholar Profile
Scientific Characteristics of Dr. Karol Grudziński for the post-doctoral time (2002-by now)
The issue of minimizing the functions of many variables is so fundamental that a lot of attention has been paid to it. The perennial problem of scientists is still not fast enough minimization methods that are a bottleneck in studying physical, mathematical problems, etc. Researchers outrun in developing and refining many methods of minimization to obtain the fastest and most accurate algorithms. One of the newer minimization methods, or rather one of the more recent modifications of the long-known simplex optimization procedure, is the EkPMinimizer system by Karol Grudziński. It is a modification of the simplex method invented by Nelder and Mead in the 60s and implemented by Martin Lampton in Java. The improvement of this implementation by Lampton, made by Karol Grudziński, consists mainly in limiting the number of points on which the simplex is spanned. If we mark the dimension of our problem with N, the modification constituting the EkPMinimizer system consists mainly in selecting M simplex points where M << N + 1.
The first system that was based on EkPMinimizer was the EkP reference vector selection system. It is a method of selecting prototypes based on compression of the training set and used to classify the test set based on this compressed (reduced) training set. The EkP method belongs to a crucial division of machine learning, and basing it on EkPMinimizer makes this algorithm allow you to classify millions of samples in the blink of an eye and with very high accuracy. PM-M (a modification of SBL-PM-M based on amoeba simplex minimizer from Numerical Recipes) developed by Karol Grudziński is the second prototype selection algorithm that has been based on the EkPMinimizer engine. Like EkP, it also benefits from EkPMinimizer’s advantages and offers incredible speed of classification and very high classification accuracy.
Karol Grudziński has developed many more very precious algorithms that are all based together on the EkPMinimizer optimization engine. With time, these algorithms will be gradually improved, reimplemented, published, and delivered to academic and business communities.
Selected Recent Publications and Papers in Preparation
I work on many things, both on new algorithms as well as on numerous unaccomplished and unpublished models.
Downloads
Nearly all my papers can be found here.
Drop instance pruning system written by Randall Wilson and modified by Karol Grudziński. This archive contains binaries and sources for Linux and Windows systems.