Data-efficient Online Classification with Siamese Networks and Active Learning

Abstract

An ever increasing volume of data is nowadays becoming available in a streaming manner in many application areas, such as, in critical infrastructure systems, finance and banking, security and crime and web analytics. To meet this new demand, predictive models need to be built online where learning occurs on-the-fly. Online learning poses important challenges that affect the deployment of online classification systems to real-life problems. In this paper we investigate learning from limited labelled, nonstationary and imbalanced data in online classification. We propose a learning method that synergistically combines siamese neural networks and active learning. The proposed method uses a multi-sliding window approach to store data, and maintains separate and balanced queues for each class. Our study shows that the proposed method is robust to data nonstationarity and imbalance, and significantly outperforms baselines and state-of-the-art algorithms in terms of both learning speed and performance. Importantly, it is effective even when only 1% of the labels of the arriving instances are available.

Similarity Learning based Few Shot Learning for ECG Time Series Classification

“…on a few shot classifications on ECG data using Siamese convolutional neural networks that can work satisfactorily without a large amount of labelled data. Siamese convolutional neural networks make use of similarity learning that helps to overcome the problem of class imbalance as ECG data are mostly imbalanced with the majority of the beats are normal, and only a few beats are abnormal [14]. We have shown that SCNNs are capable of learning feature embeddings using onedimensional filters of different lengths.…”

“…SCNN [13] learns a similarity score between pairs of time series. Since SCNN makes use of pairs of time series to learn similarity score, it is not affected by class imbalance [14], and this is one of the reasons for using SCNN in our work. In particular, we use a CNN with various filters and concatenate them to learn a feature embedding of each time series.…”

Online Learning With Adaptive Rebalancing in Nonstationary Environments

“…We have shown in Section VII-C that AREBA maintains its dual nature benefits and still outperforms other state-of-the-art algorithms in conditions where the assumption is violated. Future work will relax this assumption and examine other paradigms, such as, active learning [51].…”

Nonstationary data stream classification with online active learning and siamese neural networks✩

“…ActiQ: A state-of-the-art uncertainty-based algorithm [11] which uses the RVUS, but within the proposed architecture shown in Fig. 1.…”

“…ActiSiamese was first introduced in our brief conference paper [11], and this work constitutes a significant extension of that. Unlike our preliminary paper:…”