Abstract:In image classification, feature combination is often used to combine the merits of multiple complementary features and improve the classification accuracy compared with one single feature. Existing feature combination algorithms, e.g., multiple kernel learning, usually determine the weights of features based on the optimization with respect to some classifier-dependent objective function. These algorithms are often computationally expensive, and in some cases are found to perform no better than simple baselines. In this paper, we solve the feature combination problem from a totally different perspective. Our algorithm is based on the simple idea of combining only base kernels suitable to be combined. Since the very aim of feature combination is to obtain the highest possible classification accuracy, we measure the combination suitableness of two base kernels by the maximum possible cross-validation accuracy of their combined kernel. By regarding the pairwise suitableness as the kernel adjacency, we obtain a weighted graph of all base kernels and find that the base kernels suitable to be combined correspond to a cluster in the graph. We then use the dominant sets algorithm to find the cluster and determine the weights of base kernels automatically. In this way, we transform the kernel combination problem into a clustering one. Our algorithm can be implemented in parallel easily and the running time can be adjusted based on available memory to a large extent. In experiments on several data sets, our algorithm generates comparable classification accuracy with the state of the art.

Combination of Multiple Nearest Neighbor Classifiers Based on Feature Subset Clustering Method

Combination Of Multiple Real-Valued Nearest Neighbor Classifiers Based On Different Feature Subsets With Fuzzy Integral

Combination of Multiple K-NNCS by Fuzzy Integral

Combining KNN algorithm and other classifiers.

CMNN: Coupled Modular Neural Network.

Multiple Nearest Neighbor Algorithm Based on Data Mining

Combination of Multiple Classifier Using Feature Space Partitioning

AN IMPROVEMENT TO THE NEAREST NEIGHBOR CLASSIFIER AND F ACE RECOGNITION EXPERIMENTS

On The Use of Nonparametric Neighborhood Classification Rules in Multiple Classifier Combination

An Improvement To The Nearest Neighbor Classifier And Face Recognition Experiments

Sequential Combination Methods for Data Clustering Analysis

Fuzzy Nearest Neighbor Partitioning Neural Network for Classification

Feature Combination Via Clustering

Improvement of FCM Neural Network Classifier Using K-Medoids Clustering

Linear Regression Based Combination of Neural Classifiers.

Coarse to Fine K Nearest Neighbor Classifier

Multi-dimensional classification via kNN feature augmentation

MHCNC: A Novel Framework for Multi-Source Heterogeneous Cross-Network Node Classification

A Novel Feature Grouping Method for Ensemble Neural Network Using Localized Generalization Error Model

Modified center-based feature line classification approach

Nearest Neighbor Matching for Deep Clustering