Abstract
Numerous stochastic search methods have been applied in parameter estimation problems in genetic network identification. In this work, a constrained stochastic space search (CSSS) method for parameter estimation is proposed and used to optimize the goal function for the difference between measured and estimated gene expression time series data. Both linear and nonlinear model formalism were used. The performance of the proposed optimization method was compared to another robust stochastic algorithm(ICRS/DS), which is a modification of the ICRS algorithm [1]. Even though, the ICRS/DS method was shown to be robust, the problemwith using it is that thismethod requiresmaking heuristic guesses of various tuning parameters for initialization.
The ICRS/DS also takes a long time to achieve convergence to optimum solutions. To address these problems an alternative method (the CSSS) is introduced, a method
uses a technique of variance scaling on the parameters. This avoids the necessity to make heuristic guesses and speeds up the optimization process. The CSSS algorithmis fast and efficient when applied to less noisy time series data sets from small-sized genetic networks.
Original language | English |
---|---|
Title of host publication | 30th Benelux Meeting on Systems and Control, Lommel, Belgium, 15 - 17 March, 2011 |
Pages | 129-129 |
Publication status | Published - 2011 |
Event | 30th Benelux Meeting on Systems and Control, Lommel, Belgium - Duration: 15 Mar 2011 → 17 Mar 2011 |
Conference
Conference | 30th Benelux Meeting on Systems and Control, Lommel, Belgium |
---|---|
Period | 15/03/11 → 17/03/11 |