Applications of Probabilistic Modeling and Search Methods (PROMISE)
PROMISE is a two-year research project funded by the Technology Development Center (TEKES) as part of the national research programme on Adaptive and Intelligent Systems Applications. The budget of this two-year research project is approximately FIM 7 million (USD 1.2 million).
The goal of the PROMISE project is to develop methods for applying probabilistic modeling and optimization techniques in building and using adaptive and intelligent systems.
The project focuses on two research areas: probabilistic modeling and stochastic optimization. In probabilistic modeling, the main goal of the project is to develop computationally efficient methods for building and applying probabilistic models, such as Bayesian networks and finite mixture models. In stochastic optimization, the goal is to empirically study and compare different stochastic search methods, such as simulated annealing and genetic algorithms, in complex, highly constrained problem domains. All the methods developed in the project will be validated by using real-world problems provided by the industrial partners.
For more information, please contact:
The goal of the project is to develop methods for using background knowledge in neural network modelling. We study means for including statistical, functional and structural knowledge from the target system to the neural network models. In the project we develop Bayesian methods for combining the background knowledge and the evidence from the data, which also provides efficient tools for controling the complexity of the model. The research focuses on problems like how to map the background knowledge to the prior distribution of the network parameters, and how to efficiently evaluate the validity of the network with respect to the background knowledge.
The methodological research is done in close contact with the case problems, ensuring that the background knowledge comes from actual application domain experts and the methods are tested with real data. The case problems are related to inverse problems in process tomography, estimation of load by measuring the strains in the supporting frame, and modelling of quality parameters in concrete and forest industry.
For more information, please contact:
The project aims at developing and implementing advanced inventory control methods for the supply chain. One focus is solving nonstandard problems of demand forecasting. The forecasting needs of the pilot companies are analyzed, and solutions developed, the emphasis being on functionalities not generally found in existing commercial forecasting systems. The other focus is optimizing purchase levels and other inventory-related decisions, and making these decision processes adaptive. The needs of the pilot companies are analyzed within the framework of the whole company, and its information infrastructure (such as enterprise application software). Optimization solutions scalable to the problem size, and based on demand forecasts, are developed.
For more information, please contact: