A Multi-class Pattern Recognition Method by Combined Use of Multinomial Logit Model and K-Nearest Neighbor Rule

In this paper, we propose a method for multi-class pattern classification by combined use of Multinomial Logit Model (MLM) and K-nearest neighbor rule (KNN). Multinomial Logit Model (MLM) is one of the neural network models for multi-class pattern classification, and is supposed to be equal or better in class&cation performance than linear classification methods. K-NN is a simple but powerful non-parametric classification tool whose error probability does not exceed double of byes error. However, it is also known that such high performance of K-NN is not always expected if number of dimension of input feature vector space is large. Therefore, first we train MLM using the training vectors, and then apply K-NN to the output of the MLM. By this, since K-NN is applied to the compressed low dimension vectors, it is expected not only to bring out natural performance of K-NN but also to shorten computation time. Evaluation experiments were conducted by using some sets of non-artificial samples extracted from the handwritten character image database "ETL6". Those are (1) 36classes (number + English capital letter), and (2) 82-classes (number + English capital letter + "Katakana" letter). Consequently, we obtained the following recognition rates: (1) 36classes =. 100.0%, and (2) 82-classes =. 99.93%.