A neural computer architecture for data exploration

The authors´ approach to problem-solving does not require a formal, symbolic mathematical model. It is based on interactive experimentation, interrelating the result data into IF-THEN units, data generalization and data-driven reasoning. It can be applied to the exploration of real-word hard-to-formalize data-intensive problems such as traffic control, environmental forecasting, military reconnaissance, technical and medical diagnosis, robotics, software debugging, and system maintenance. The data-based model, in contrast to a symbolic one, can be stepwise refined using additional input-output pairs to match any given precision or other constraints, a possibility hard to realize in the case of symbolic mathematical models