Contributions and Results obtained

The contributions of the research reflected in this thesis can be summarized as follows.

• An integrated environment to serve as a possible assistant during the educational process was built, which I called Subsymbolic Knowledge Extraction Environment (SSKEE).
• I developed the three modules (or building blocks) of the SSKEE: the neural network module, the knowledge extraction module and the user interface module.
• A new energy function for supervised learning on feedforward networks, based on restrictions was proposed and was used in the neural network module.
• The new energy function was tested separately against some benchmarks and with the help of some coefficients and was proved to be less sensitive to minor errors when compared with the standard backpropagation function. The successive results of these tests were presented at several national and international conferences (see, e.g., [31, 22, 24, 25, 26, 28, 29]) and an integrated description was accepted for publication in an international journal ([17]).
• I have built a neural network for stock exchange prediction based on this energy function. This work is reflected in another paper written for an international journal, which accepted it for publication (see [18]).
• The knowledge extraction module was developed. For testing, a case study of teaching stock exchange occurrences and in particular stock exchange price evolution in a regular Economy class was used. I reported the results briefly at some international conferences (see [23] and [27]) and at length in an international journal ([16]).
• In the context of the ANN implementation, the parallel aspects of an ANN, as well as the advantages of a parallel implementation of a ANN were considered and studied. The different levels of parallelism, namely coarse, medium and fine grain parallelism are put in correspondence with the ANN functions and structure. For instance, at a coarse grain parallelism, the layer parallelism is considered. At the medium level, I studied the neuron level parallelism. At the fine level I had to consider the very functions of the ANN and to enter the instruction level parallelism. Out of these types, an original combined optimization is built and described. The results of this line of research were published at some international conferences ([20], [21]) and a longer version in an international journal ([19]).