Workpackage : WP2
Applications of Neural Networks - collaboration and research visits
Start date or start event : 1 month
Objectives :
Collaboration and exchange of researchers developing pioneer problems in computer physics:
  1. Applications of the Neural Network techniques for the optimisation of multidimensional problems.
  2. Exploring genetic algorithm principles to tune parameters of the neural network.
  3. Exploring parallel learning process.
  4. Developing universal separation method which can be used for various pattern recognition problems.
Description :
Neural networks are broadly used in pattern recognition problems. There are many types of artificial neural networks which differ in architecture, in the type of implemented transfer functions and strategy of learning. Regard to universal approximation property we used for pattern recognition problem a special kind of neural nets, namely neural network with switching units. This neural network is a combination of classical neural network architecture and decision tree. The merit of the work lies in the application of Genetic Algorithms (GA) procedures of crossover and mutation to find better topology and parameters of neural networks. To improve overall performance parallel version of neural net learning process is also implemented. Programme uses Parallel Virtual Machines library. The above problems are a separated and well defined part of broader European collaboration preparing the high energy experiments at CERN and DESY-Hamburg. The results of research are disseminated to undergraduate and graduate students in the form of lectures.
Number of visits : Partners involved :
Deliverables :
  1. Report on the status of collaboration and obtained results
  2. Lecture course for advanced undergraduate and graduate students
Milestones and expected results :
  1. Establishment of close collaboration between three research centres.
  2. Common publications in renowned journals.
  3. Strengthening and integration of these research centres with the European Research Area.