Model-free deconvolution of transient signals using genetic algorithms


METADATA ONLY
Loading...

Date

2012-03

Publication Type

Book Chapter

ETH Bibliography

no

Citations

Altmetric
METADATA ONLY

Data

Rights / License

Abstract

Model-free deconvolution of transient signals usually suffers from a wavy small-frequency component as well as an increase of high frequency noise. The more these features are damped, the more the deconvolved signal becomes distorted. If we can include as much information concerning the genuine deconvolved signal as possible, the quality of the deconvolved transient can be greatly improved. We have developed a deconvolution procedure which does not make use of a presupposed (often arbitrary) functional form of the transient signal. Instead, it is based on the inversion of the distortion effects of convolution, which means temporal compression, amplitude enhancement, increasing of the steepness of rise and decay of the measured convolved signal, and cutting initial data to zero to reproduce an eventual sudden jump. As our results have proved that the choice of a fairly good initial population is crucial for a successful deconvolution, we use additional genetic algorithms to generate an initial population with a relatively high fitness. The initial population is generated from the measured convolved signal in two subsequent stages, each consisting of a genetic algorithm and the check of the potential to improve the fitness of the population during further breeding, as a candidate for the deconvolved data set. This generation can be done “automatically” so that the user does not need to experiment much to get a good initial population and satisfactory final results. The final genetic algorithm performing the main iteration to get the estimate of the deconvolved data is constructed in such a way that it does not enhance either the low-frequency wavy behaviour, or the high frequency noise. It is based on a smooth mutation in a range containing several data, and a dynamic adjustment of the mutation. The method is described in connection with the deconvolution of ultrafast laser kinetic (or femtosecond chemistry) data. The advantage of this method to find an underlying molecular mechanism via statistical inference is also discussed.

Permanent link

Publication status

published

Book title

Handbook of Genetic Algorithms: New Research

Journal / series

Volume

Pages / Article No.

41 - 59

Publisher

Nova Science Publishers

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Deconvolution; Femtochemistry; Fluorescence decay; Genetic algorithm; Transient signals

Organisational unit

09759 - Bezdek, Máté József / Bezdek, Máté József check_circle

Notes

Funding

Related publications and datasets