On the Nonlinearity of Pattern Classi
نویسنده
چکیده
This paper presents a novel approach to the analysis of the overtraining phenomenon in pattern classiiers. A nonlinearity measure N is introduced which relates the shape of the classiication function to the generalization capability of a classiier. Experiments using the k-nearest neighbour rule, a neural classiier and the quadratic classiier show that the introduced measure N can be used to study the overtraining behaviour of a classiier. Moreover N shows to be a predictor for the local sensitivity of a classiier. Classiiers that have a small local sensitivity are shown to have a low nonlin-earity, whereas an increased nonlinearity indicates an increase in local sensitivity.
منابع مشابه
Mathematically modeling of target speed effect on nonlinearity in DTLMI-based nano-metrology system
Jones matrix computation is one of the widely used methods in nonlinearitycalculation in laser interferometers. In this paper, the nonlinearity error in developedthree-longitudinal mode heterodyne interferometer (DTLMI) has been mathematicallymodeled at various speeds, by using Jones matrix calculations. This review has been donedespite the fact that simultaneously the main factors including no...
متن کاملA Note on Free Vibration of a Double-beam System with Nonlinear Elastic Inner Layer
In this note, small amplitude free vibration of a double-beam system in presence of inner layer nonlinearity is investigated. The nonlinearity is due to inner layer material and is not related to large amplitude vibration. At first, frequencies of a double-beam system with linear inner layer are studied and categorized as synchronous and asynchronous frequencies. It is revealed that the inner l...
متن کاملEmploying Foundation Nonlinearity to Mitigate Seismic Demand in Superstructure
Because of difficulty in inspection and retrofit of foundation in comparison with other elements, the common design philosophy is to avoid any nonlinear deformation in the foundation. This paper shows that by employing controlled foundation nonlinearity, in predetermined sections with arrangements for inspection and retrofit, it is possible to reduce seismic demand on superstructure. Localizing...
متن کاملOn convergence of homotopy analysis method to solve the Schrodinger equation with a power law nonlinearity
In this paper, the homotopy analysis method (HAM) is considered to obtain the solution of the Schrodinger equation with a power law nonlinearity. For this purpose, a theorem is proved to show the convergence of the series solution obtained from the proposed method. Also, an example is solved to illustrate the eciency of the mentioned algorithm and the h-curve is plotted to determine the region ...
متن کاملHierarchical classification and feature reduction for fast face detection with support vector machines
We present a two-step method to speed-up object detection systems in computer vision that use support vector machines as classi ers. In the rst step we build a hierarchy of classi ers. On the bottom level, a simple and fast linear classi er analyzes the whole image and rejects large parts of the background. On the top level, a slower but more accurate classi er performs the nal detection. We pr...
متن کامل