A Recursive Ensemble Learning Approach With Noisy Labels or Unlabeled Data

Abstract

For many tasks, the successful application of deep learning relies on having large amounts of training data, labeled to a high standard. But much of the data in real-world applications suffer from label noise. Data annotation is much more expensive and resource-consuming than data collection, somewhat restricting the successful deployment of deep learning to applications where there are very large and welllabeled datasets. To address this problem, we propose a recursive ensemble learning approach in order to maximize the utilization of data. A disagreement-based annotation method and different voting strategies are the core ideas of the proposed method. Meanwhile, we provide guidelines for how to choose the most suitable among many candidate neural networks, with a pruning strategy that provides convenience. The approach is effective especially when the original dataset contains a significant label noise. We conducted experiments on the datasets of Cats versus Dogs, in which significant amounts of label noise were present, and on the CIFAR-10 dataset, achieving promising results.

Harnessing artificial intelligence for the next generation of 3D printed medicines

“…In the modern world of big data, partially labelled datasets are a common occurrence [170]. Unsupervised learning techniques are used to label unlabelled data by drawing inferences from data with labels [171]. Subsequently, supervised techniques are then used to identify relationships between data features and their labels.…”

A Fundamental Overview of SOTA-Ensemble Learning Methods for Deep Learning: A Systematic Literature Review

“…This is exactly where Ensemble Learning (EL) comes into play. EL is a way to solve DL problems by combining several individual models that provide independent results and aggregating them to make a final decision based on consensus [1], [9], [10], [14]. EL is thus an effective way to optimize generalization, predictive performance, and robustness through a combined model [1], [10], [11], [14]- [16].…”

“…The EL is built on the same principle: a common decision is made based on several individual decisions of models [9], [18], [19]. In the context of DL, an ensemble can be defined as a ML system composed of several individual models that learn in parallel or sequentially [1], [9], [10], [14], [17], [19]- [21]. The individual results are combined to find a collective solution to a specific problem [1], [9], [10], [14], [17], [19]- [21].…”

Hybrid Dimensionality Reduction Forest With Pruning for High-Dimensional Data Classification

“…Pratama et al [31] propose a novel evolutionary classifier ensemble method, called parsimonious ensemble, which achieves a tradeoff between precision and complexity. Wang et al [32] propose a recursive ensemble learning approach to maximize the use of data in deep learning applications. He and Cao [33] propose a classifier combination method based on signal strength, which combines the output of multiple classifiers to assist decision-making.…”