An Incremental Hypersphere Learning Framework for Protein Membership Prediction



With the recent raise of fast-growing biological databases, it is essential to develop efficient incremental learning algorithms able to extract information efficiently, in particular for constructing protein prediction models. Traditional inference inductive learning models such as SVM perform well when all the data is available. However, they are not suited to cope with the dynamic change of the databases. Recently, a new Incremental Hypersphere Classifier (IHC) Algorithm which performs instance selection has been proved to have impact in online learning settings. In this paper we propose a two-step approach which firstly uses IHC for selecting a reduced data set (and also for immediate prediction), and secondly applies Support Vector Machines (SVM) for protein detection. By retaining the samples that play the most significant role in the construction of the decision surface while removing those that have less or no impact in the model, IHC can be used to efficiently select a reduced data set. Under some conditions, our proposed IHC-SVM approach is able to improve performance accuracy over the baseline SVM for the problem of peptidase detection.


Incremental Learning, Support Vector Machines, Incremental Hypersphere Classification


Incremental learning

Related Project

BIOINK- Incremental Kernel Learning for Biological Data Analysis


International Conference on Hybrid Artificial Intelligence Systems, LNCS 7208, pp. 429-439, March 2012


Cited by

Year 2013 : 1 citations

 Chen, Y. H., et al. "A Hybrid Text Classification Method Based on K-Congener-Nearest-Neighbors and Hypersphere Support Vector Machine." Information Technology and Applications (ITA), 2013 International Conference on. IEEE, 2013.