模式识别prch5part3_ding.ppt

Pattern ClassificationAll materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John Wiley Sons, 2000 with the permission of the authors and the publisher;5.8 Minimum Squared Error Procedures;Minimum Squared Error and Pseudoinverse For all the samples we want a weight vector a so that for some arbitrarily specified positive numbers. The matrix notation : Error vector:;Sum-of-squared-error criterion function: The gradient Set it to zero, we get If is nonsingular, The d by n matrix is call the pseudoinverse of Y. Remarks: For an arbitrarily fixed b, MSE solution may not be a separating vector. ;;The Widrow-Hoff or LMS Procedure Need not converge to a separating hyperplane even if there exist one;5.9 The Ho-Kashyap Procedure;鸯梭丫仆驮却绦督律钾属办喉敷镶雨善生施蹬符秦拜灯丘汽郁跟象趣滇衰模式识别prch5part3_ding模式识别prch5part3_ding;5.12 Multicategory Generalizations;Kesler’s Construction suppose that , we have ;Remark: it allows us to convert many muylticategory error-correction procedures to two-category procedures for the purpose of obtaining a convergence proof Fixed-Increment Rule Let denote the kth sample requiring correction, and suppose that , there must be at least one j i for which then the fixed-increment rule for correcting is ;Remark:Kesler’s construction can also be used to extend all of results for the Perceptron and relaxation procedure to the multicategory case;Generalization for MSE Procedure consider multicategory case as a set of c two-class problem