Probabilistic distribution sampling operation with a use of Mathematica (http://www.wolfram.com/mathematica/) software. org.embl.ebi.escience.scuflworkers.java.XMLInputSplitter org.embl.ebi.escience.scuflworkers.java.XMLOutputSplitter org.embl.ebi.escience.scuflworkers.java.XMLOutputSplitter http://localhost:8080/tav4sb/services?wsdl sampleDistribution source(new URL("http://localhost/tav4sb/workflows/setDistributionXml.bsh")); name parameterList distributionXml head = list.get(0); tail = list.subList(1, list.size()); list head tail out = in; in out Desired number of samples. Note: for the "full" method the resulting number of samples will be equal to number of samples to the power equal to dimension of each sample. List of lists which specify probabilistic distributions to sample from. Size of this list defines dimension of samples. Each distribution is specified by name and parameter values. Name is a string with a prefix from Mathematica's *Distribution symbols. See the Mathematica's documentation for comprehensive list o available distributions and their parameters: http://reference.wolfram.com/legacy/v7/guide/StatisticalDistributions.html . Parameters are specified using number, string, Mathematica symbol or heterogeneous list of any of those types . These are represented as: d.n "text" Symbol { el1, el2, ... } repectively, i.e. text containing '.' character is treated as a decimal number (e.g. '1.0', '0.15', etc.), quoted text is interpreted as a string, plain text w/o '.' character is interpreted as a Mathematica symbol and curly brackets delimit list, which elements are comma-separated. ATTENTION: comas and quotes inside of the string must be escaped with backslash (i.e. "\," or "\""); you can also escape curly brackets for consistency but it's not mandatory. Example - two dimensional samples defined by two lists of three and two elements respectively: "normal", 0.0, 1.0 "uniform", {0.0, 1.0} Method is one of: * "random" (default) - standard, random distribution sampling, * "lhs" - Latin Hypercube sampling (see http://en.wikipedia.org/wiki/Latin_hypercube_sampling), * "full" - "lhs" over each of the sampled dimensions with theirs full product, i.e. each grid part of the sample space is represented by one sample; note: returned number of samples is a given number of samples to the power equal to dimension of a sample.