Abstract:
A Self Organizing Feature Map (SOFM) is a kind of neural architecture making an acceptable compromise between information complexity, and the need of explanatory capabilities for an assigned system. Such Properties make SOFM a powerful support to be used over specific contexts (financial markets,..), where mathematical modelling efforts have often been vanished, and no alternative choices are offered to researchers, but for technical tools, the orthodoxity of which is still far to be completely accepted. SOFM most relevant property relies, perhaps, in learning procedure flexibility. On the other hand, we have to face against such a structural rigidity affecting the map, because its planar dimensions are settled during earlier steps of the process. This evident bottleeck of the procedure lead us to develop an algorithm which overcomes those limits by combining the approach of SOFM to that proper of Evolutional Models. The result is a process integrating knowledge from Self Organization Theory and Genetics (ISOG), the efficiency of which has been tested on financial markets.