Deep stacked stochastic configuration networks for lifelong learning of non-stationary data streams

Pratama, Mahardhika; Wang, Dianhui

doi:10.1016/j.ins.2019.04.055

Cited by 57 publications

(16 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is done by associating each data entity with a rumour index, which reflects its current state of being influenced by harmful information. Measuring falsehood is essentially a problem of probability modelling, which compresses detective cues into a probability value, as studied in Bayesian statistics, frequentist statistics, and machine learning [1,21]. Several detective cues have been investigated [4,8,22]:…”

Section: Streaming Social Signalsmentioning

confidence: 99%

JUDO: Just-in-time rumour detection in streaming social platforms

Nguyen

Nguyên

Nguyen

et al. 2021

Information Sciences

View full text Add to dashboard Cite

Web platforms, especially social media, are facing a new and ever-evolving cyber threat operating at the information level. Their open nature allows a high velocity flow of rumours that emerge unexpectedly and spread quickly. While rumour detection has attracted many theoretical and practice studies, the timing of the detection is often neglected or not properly considered. Rumours often cause irreversible damage worldwide before being successfully detected. To address this, we approach early rumour detection from a streaming perspective. We present a just-in-time rumour detection framework that is built on top of the continuous scoring of rumour-related signals. To overcome the trade-off between timeliness and the coefficient of detection, our model treats social graphs as a data stream and computes the anomaly score of potential rumours at both the element-level and subgraph-level. This multi-level approach not only captures the propagation structure of rumours but also focuses on abnormal elements that are responsible for bootstrapping or amplifying the rumours (the 'explore vs exploit' effect). With extensive evaluations on our published benchmark, our model identifies rumours earlier than the baselines while achieving an even better detection coefficient.

show abstract

Section: Streaming Social Signalsmentioning

confidence: 99%

JUDO: Just-in-time rumour detection in streaming social platforms

Nguyen

Nguyên

Nguyen

et al. 2021

Information Sciences

View full text Add to dashboard Cite

show abstract

“…This section discusses experimental study of DEVFNN in five popular real-world and synthetic data stream problems: electricity-pricing, weather, SEA, hyperplane, SUSY, and indoor RFID localization problem from our own project. DEVFNN is compared against prominent data stream algorithms in the literature: pENsemble [32], pENsemble+ [34], DSSCN [33], gClass [35]. Comparison with pENsemble and pENsemble+ are presented to illustrate the advantages of deep stacked network structure against ensemble structure, while DSSCN represents a deep algorithm created by the concept of random shift rather than the feature augmentation approach.…”

Section: Numerical Studymentioning

confidence: 99%

“…A novel incremental DNN, namely Deep Evolving Fuzzy Neural Network (DEVFNN), is proposed in this paper. DE-VFNN features a fully elastic structure where not only its fuzzy rule can be autonomously evolved but also the depth of network structure can be adapted in the fully automatic manner [33]. This property is capable of handling dynamic variations of data streams but also delivering continuous improvement of predictive performance.…”

Section: Introductionmentioning

confidence: 99%

An Incremental Construction of Deep Neuro Fuzzy System for Continual Learning of Non-stationary Data Streams

Pratama

Pedrycz

Webb

2019

IEEE Trans. Fuzzy Syst.

Self Cite

View full text Add to dashboard Cite

Existing fuzzy neural networks (FNNs) are mostly developed under a shallow network configuration having lower generalization power than those of deep structures. This paper proposes a novel self-organizing deep fuzzy neural network, namely deep evolving fuzzy neural networks (DEVFNN). Fuzzy rules can be automatically extracted from data streams or removed if they play little role during their lifespan. The structure of the network can be deepened on demand by stacking additional layers using a drift detection method which not only detects the covariate drift, variations of input space, but also accurately identifies the real drift, dynamic changes of both feature space and target space. DEVFNN is developed under the stacked generalization principle via the feature augmentation concept where a recently developed algorithm, namely Generic Classifier (gClass), drives the hidden layer. It is equipped by an automatic feature selection method which controls activation and deactivation of input attributes to induce varying subsets of input features. A deep network simplification procedure is put forward using the concept of hidden layer merging to prevent uncontrollable growth of input space dimension due to the nature of feature augmentation approach in building a deep network structure. DEVFNN works in the sample-wise fashion and is compatible for data stream applications. The efficacy of DEVFNN has been thoroughly evaluated using six datasets with non-stationary properties under the prequential test-then-train protocol. It has been compared with four state-ofthe art data stream methods and its shallow counterpart where DEVFNN demonstrates improvement of classification accuracy. Moreover, it is also shown that the concept drift detection method is an effective tool to control the depth of network structure while the hidden layer merging scenario is capable of simplifying the network complexity of a deep network with negligible compromise of generalization performance.

show abstract

“…In such scenarios, the fixed learner obtained by training deep networks on pre-collected samples always cannot finely adapt the varying environment, and the over-fitting issue then inevitably appears in the testing stage. Thus it is also a critical issue to make deep learning strategy automatically adaptable to open and dynamic environments [42,43] . How to improve the generalization capability of a deep learning approach should be the most significant issue for attaining this "learn to adapt" purpose.…”

Section: Automatic Adapting To Dynamic Environmentsmentioning

confidence: 99%