Pattern Recognition of Multivariate Information Based on Non-Statistical Techniques

A novel method for non-statistical PR/dimensionality reduction technology of multivariate information is proposed, which based on graphical techniques, quantitative non-statistical technique, grey system theory, fuzzy system theory, Information entropy theory, neronet technique et al. In nowadays pattern recognition field, statistical pattern recognition techniques are exercising dominion over. But there are many real-world problems the sample is not obey the known Statistical models. And it is Sometimes difficulty and costliness to obtain information, multivariate and small sample problems should be solved in PR techniques. All of above can't be faultlessly solved by statistical approaches. Non-statistical pattern recognition methods provide an interesting alternative to statistical approaches, and must be employed for problems that cannot be handled using statistical methods. Applied the non-statistical techniques in PR technology, we can avoid the distortion and aberration of information because the statistical condition was insufficiency and improve the reliability of PR program study. Applying the non-statistical techniques in our data experiment of multivariate information feature selection, we achieved better performance. We believe that this method can be used in many other multivariate information dimensionality reduction methods, and will obtain better performance than them.