[Machine Learning | Incremental Learning] Learn++.NC: Combining Ensemble of Classifiers With Dynamically Weighted Consult-and-Vote for Efficient Incremental Learning of New Classes

Overview / 概览

This paper introduce an incremental learning algorithm called Learn++ .NC (New class). It's an upgrade version of Learn++. Learn++ use a two layered structure handle incremental learning problem. The first layer is weak-learner, each time a new data branch comes in, Learn++ will training a new set of weak-learner. Second layer is linear voting with uniform weight.

本片论文介绍了一个名为Learn++ .NC 的增量学习算法, 该算法为Learn++ 的升级版. Learn++ 使用了两层结构来解决增量学习的问题. 第一层是弱学习层, 该层包含多个弱学习器, 每当有新数据集到来, Learn++ 将会训练一批新的弱学习器. 第二层为一个统一权重的线性投票层.

But Learn++ has some weakness, such as "outvoting". Base on the paper said ""

[1] Muhlbaier, M., Topalis, A. and Polikar, R. (2009). Learn++.NC: Combining Ensemble of Classifiers With Dynamically Weighted Consult-and-Vote for Efficient Incremental Learning of New Classes. IEEE Transactions on Neural Networks, 20(1), pp.152-168.