Fuzzy Logic Evaluation of Soybean Plant Shape.

Combining Regression Trees and Radial Basis Function Networks

“…Not all features are independent, some are combinations of the others, for example D = AREA / WDT HGT. A variety of learning methods have previously been used on this data set, including linear discriminant analysis Ninomiya and Shigemori 1991], fuzzy logic Ambuel et al, 1997] and classi cation trees Ninomiya and Nguyen-Cong, 1998]. Oide and Ninomiya 1998] used multilayer perceptrons but with the original images, rather than feature vectors, as the inputs.…”

“…These can be compared to similar tables in Ambuel et al 1997] and Ninomiya and Nguyen-Cong 1998] though di erences in compilation should be noted. In Ambuel et al 1997], results are quoted for two methods (fuzzy logic and linear discriminant analysis) and applied to a superset of our data (875 cases).…”

“…These can be compared to similar tables in Ambuel et al 1997] and Ninomiya and Nguyen-Cong 1998] though di erences in compilation should be noted. In Ambuel et al 1997], results are quoted for two methods (fuzzy logic and linear discriminant analysis) and applied to a superset of our data (875 cases). After renormalising the results in Ambuel et al 1997] to our smaller data set (326), it is apparent that both these methods are considerably less accurate than the method described here.…”

“…In Ambuel et al 1997], results are quoted for two methods (fuzzy logic and linear discriminant analysis) and applied to a superset of our data (875 cases). After renormalising the results in Ambuel et al 1997] to our smaller data set (326), it is apparent that both these methods are considerably less accurate than the method described here. However, it should be borne in mind that the extra data in Ambuel et al 1997] consisted of cases where the human experts were not unanimous in their judgements of the correct classi cation.…”

“…After renormalising the results in Ambuel et al 1997] to our smaller data set (326), it is apparent that both these methods are considerably less accurate than the method described here. However, it should be borne in mind that the extra data in Ambuel et al 1997] consisted of cases where the human experts were not unanimous in their judgements of the correct classi cation. This may have degraded the results.…”