A Survey on Ensemble Learning for Data Stream Classification

By Heitor Murilo Gomes, Jean Paul Barddal, Fabrício Enembreck, Albert Bifet. 2017

576 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Han-Jia Ye
2. De-chuan Zhan
3. Da-Wei Zhou
4. Qiwen Wang
5. Ziwei Liu
6. Fu-Yun Wang
7. Peng Zhao
8. Zhiyuan Qi

In Signal Processing

1. Peng Zhao
2. Zhi-Hua Zhou
3. Bo-Jian Hou
4. Barbara Hammer
5. Hui Xiong
6. Yang Yang
7. Heiko Wersing
8. De-chuan Zhan

Citations in the corpus

• Few-Shot Class-Incremental Learning by Sampling Multi-Phase Tasks (2022, background): “However, real-world applications often face the stream format data [7] and novel classes [8], [9], [10], [11], [12], e.”
• Forward Compatible Few-Shot Class-Incremental Learning (2022, background): “However, data often come in stream format [18] with emerging new classes [12, 63, 65] in real-world applications, e.”
• Cost-Effective Incremental Deep Model: Matching Model Capacity With the Least Sampling (2023, unknown): “Concept drift caused by the distribution evolution is a well-recognized research direction in incremental learning [11], [30].”
• A Model or 603 Exemplars: Towards Memory-Efficient Class-Incremental Learning (2022, background): “In the open world, training data is often collected in stream format with new classes appearing [16, 15, 63, 65].”
• AI for Next Generation Computing: Emerging Trends and Future Directions (2022, unknown): “, to save resources on the edge, the data-stream model is etrained only when its performance crosses below a threshold [189].”
• Learning With Feature Evolvable Streams (2017, background): “We want to emphasize that though concept-drift happens in streaming data where the underlying data distribution changes over time [17], [18], [19], the number of features in concept-drift never changes which is different from our problem.”
• Mitigating Concept Drift via Rejection (2018, background): “A number of approaches have been proposed for learning in non-stationary environments in the presence of concept drift, whereby several recent technologies are also suited for heterogeneous types of drift [10, 11, 12, 13, 14].”
• Adaptive Deep Models for Incremental Learning: Considering Capacity Scalability and Sustainability (2019, background): “Concept drift caused by the distribution evolution is a well-recognized research direction in incremental learning and has wide applications [12, 14].”
• Class-Incremental Learning: A Survey (2023, background): “The typical training process of a deep network requires pre-collected datasets in advance, e.g. , large-scale images or texts — the network undergoes the training process of the pre-collected dataset multiple epochs.”
• Ensemble learning: A survey (2018, background): “Recently, several comprehensive surveys on ensemble learning have been published in the literature (Gomes, Barddal, Enembreck, & Bifet, 2017; Kulkarni & Sinha, 2013; Rokach, 2016; Wo ź niak, Graña, & Corchado, 2014).”
• Learning with Feature and Distribution Evolvable Streams (2020, background): “Another important category falls into the ensemble based approaches (Zhou, 2012; Gomes et al., 2017), where they adaptively add or drop base classifiers learning from historical data and dynamically adjust weights when dealing with new coming data items (Kolter & Maloof, 2005; Elwell & Polikar,…”
• Deep Class-Incremental Learning: A Survey (2023, background): “However, training data is often with stream format in the open world [6], [7].”