Cell algorithms with data inflation for non-parametric classification

The k-nearest neighbor (k-NN) classifier represents one of the most popular non-parametric classification tools. Its main drawback is the computational cost required during the search for the nearest neighbors. In this paper, we propose using two cell algorithms with data inflation as tools capable to achieve interesting tradeoffs between classification error and computational cost. The performances of the proposed algorithms are assessed experimentally on the basis of a multisensor remotely sensed image and a pen-based handwritten digit data set.