Efficient Implementation of $k$-Nearest Neighbor Classifier Using Vote Count Circuit
Authors:
- Rongshan YuRYRongshan Yu
- Wenyu JiangWJWenyu Jiang
- Wenxian YangWYWenxian Yang
- 24 April 2014
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 61 (6) , 448-452
Abstract
The k-nearest neighbor (k-NN) classification is a nonparametric method to classify objects based on the training set. It is an instance-based classifier operating on the assumption that the unknown instance is related to the known ones according to some distance/similarity functions. In this brief, a hardwareassisted algorithm, i.e., vote count, is introduced to approximate the k-NN classifier to provide a low-cost classification solution. It is found that this hardware-assisted solution achieves similar performance as that of the k-NN classifier. In addition, it is highly scalable with respect to the training sample size, which is essential for the k-NN algorithm to deliver its full potential for real-life classification problems.Keywords
This publication has 8 references indexed in Scilit:
- An Algorithm and Hardware Design for Very Fast Similarity Search in High Dimensional Space2010
- 80 Million Tiny Images: A Large Data Set for Nonparametric Object and Scene RecognitionIEEE Transactions on Pattern Analysis and Machine Intelligence, 2008
- Locality-sensitive hashing scheme based on p-stable distributions2004
- A new k-winners-take-all neural network and its array architectureIEEE Transactions on Neural Networks, 1998
- Gradient descent learning of nearest neighbor classifiers with outlier rejectionPattern Recognition, 1995
- Analog circuit implementation and learning algorithm for nearest neighbor classifiersPattern Recognition Letters, 1994
- Bias of Nearest Neighbor Error EstimatesIEEE Transactions on Pattern Analysis and Machine Intelligence, 1987
- Nearest neighbor pattern classificationIEEE Transactions on Information Theory, 1967