Solving scientific inference problems in complex systems.
Our world unique scientific modelling and inference engine, the Minerva framework, can be used to build small and complex scientific models in a modular fashion. After defining the so called forward/generative model, inference on the model parameters is fully automatic and can be done by least squares, maximum likelihood or using full Bayesian inference.
Scientific models in large scale experiments usually depend on a number of legacy programs written in different languages, for calculating aspects of the modelled system. We integrate these codes in the Seed eScience scientific cloud, after which they can be easily called over the network, automatically parallelised, and modularly combined to form very large models
The combination of the Minerva inference system and the modularisation of legacy programs on the seed eScience scientific cloud, creates a unique infrastructure for building large modular scientific models, in which all scientific assumptions, model uncertainties, and statistical distribution of observational data, are fully handled and utilised for optimal inference on model parameters.