This thesis is organized as follows.
As my research is centered arround Artificial Neural Networks, section 2 gives a brief introduction to the field of Artificial Neural Networks. I explain why I chose to work in this field, what the advantages and disadvantages are and what is to be expected.
Chapters 3 to 8 try to cover, discuss and exemplify the two aspects of optimizations of the learning process of Artificial Neural Networks: algorithmical, mathematical optimizations on one hand, and parallelization on the other.
Chapters 11 and 12 deal with extraction of this knowledge accumulated in the ANN through learning (training).
The last sections contain discussions, conclusions and further guidance for researchers.
More details are given in the following.
Chapters 3 and 4 are grouped together as a part, containing the two aspects, theoretical and practical, of Time-Series prediction.
Chapter 3 introduces the concept of Time-Series and Time Series prediction.
Chapter 4 deals with the particular time series application of stock exchange prediction. A new energy function is presented and a neural network is built based on this new energy function. The results of the prediction are discussed and comparison with benchmarks is performed.
Chapters 7 and 8 are grouped together as a part, containing the two aspects, theoretical and practical, of Parallel Neural Networks.
Chapter 7 discusses the parallel aspects of neural networks.
Chapter 8 discusses how these parallel aspects can be integrated into a neural network and what speed-up results were obtained.
Chapters 11 and 12 are grouped together as a part, containing the two aspects, theoretical and practical, of Knowledge Extraction from Neural Networks.
Chapter 11 introduces the Knowledge Extraction concept from Neural Networks and presents some knowledge extraction tools.
Chapter 12 shows how knowledge extraction can be integrated into the system and deals with an application of stock exchange teaching at an Economy class.
Conclusions are drawn at each application section (sections 4, 8 and 12), stating and summarizing the achieved results.
Finally, in part iv I draw general conclusions concerning the whole work for this thesis, give some concluding remarks and present some recommendations for future work and developments.