By leaving the options field blank you will get the default method-specific parameters.
By adding the following statements
smoothingParameter = 0.2
randomStarts_ = 60 1
randomSeed = 1
you can modify the parameters.
Our own algorithm presented in Gennemark and Wedelin (2007, PMID:17441553) is based on an efficient heuristic search for model selection, where the structure is built incrementally. Parameters are mostly estimated for one equation at a time using standard methods with local simulation (of a single equation). Sometimes parameter estimation is done using global simulation. There are three method specific parameters in our algorithm:
The smoothing parameter. The smoothing parameter affects the smoothing of the time-series data. Each time-series is smoothed individually. For each time-series the smoothing parameter is multiplied (automatically in the program) by the number of time-points in the time-series. The resulting value corresponds to the parameter s in the cubic smoothing spline metod by de Boor, C. (A practical guide to splines. Springer-Verlag, New York, 1978. p. 235-43, netlib.org). de Boor finds the cubic smoothing spline f to given data (x(i),y(i)), i=1,...,npoint, which has as small a second derivative as possible, while
s(f) = sum( ((y(i)-f(x(i)))/dy(i))^2 , i=1,...,npoint ) <= s
The default value of the smoothing parameter is 10E-8 for perfect data and 0.2 for noisy data.
To modify this parameter, input in the text area of optional parameters:
smoothingParameter = value
followed by return for a new line.
For example, if you want to set the smoothing parameter to 0.5, input:
smoothingParameter = 0.5
followed by return for a new line.
Number of random starts in the parameter estimation. In order to avoid local maxima and thereby improve stability, several random starting points can be evaluated in our parameter estimation method. There are two major steps involved:
(step 2a) Make a rough estimate of the parameters in the equation with the derivative method.
(step 2b) Improve the estimate of the parameters in the equation with local simulation based error calculation (using standard parameter estimation subroutines).
For systems with chemical rate equations, the default number of random starts in step 2a is 30 for perfect data and 60 for noisy data. In step 2b, the default number of random starts is 1, and increased to 10 for problems with extremely small amounts of data, defined as
#experiments * #points_per_time_series < 5 * #variables
For S-systems, the default number of random starts in step 2a of the parameter estimation is 4 for perfect data and 8 for noisy data. Step 2b as above.
To modify these parameters, input in the text area of optional parameters:
randomStarts_ = integerStep2a integerStep2b
followed by return for a new line. Please note the underscore!
For example, if you want to set the number of random starts in step 2a to 15 and in step 2b to 2, input:
randomStarts_ = 15 2
followed by return for a new line.
Random seed. The random seed is used in our own algorithm in order to initialize the random number generator. For problems with very few data, many similar models are reasonable (they give similar error), and the variability in found solutions between different runs with different random seeds will generally increase. The default value is 1.
To modify this parameter, input in the text area of optional parameters:
randomSeed = integerValue
followed by return for a new line.
For example, if you want to set the random seed to 4, input:
randomSeed = 4
followed by return for a new line.