Ensemble deep learning: A review

Ganaie, M. A.; Hu, Minghui; Tanveer, Mohammad; Suganthan, Ponnuthurai Nagaratnam

doi:10.1016/j.engappai.2022.105151

Cited by 755 publications

(354 citation statements)

References 181 publications

(188 reference statements)

Supporting

Mentioning

233

Contrasting

Unclassified

Order By: Relevance

“…As stated in the Introduction, deep ensemble learning is traditionally defined as building an ensemble of multiple predictions originating from different deep convolutional neural network models [23]. However, recent novel techniques necessitate redefining ensemble learning in the deep learning context as combining information, most commonly predictions, for a single inference.…”

Section: Ensemble Learning Techniquesmentioning

confidence: 99%

“…We excluded the Boosting technique, which is also commonly used in general ensemble learning. The reason for this is that Boosting is not feasibly applicable for image classification with deep convolutional neural networks due to the extreme increase in training time [23], [25]. An overview diagram of the four techniques can be seen in Figure 2.…”

Section: Ensemble Learning Techniquesmentioning

confidence: 99%

“…In contrast to single algorithm approaches, the ensemble of different deep convolutional neural network architectures (also called inhomogeneous ensemble learning) showed strong benefits for overall performance [10], [23], [25], [26], [62]. This kind of ensemble learning is more complex and can consist of even different computer vision tasks [10], [23], [26]. The idea of the Stacking technique is to utilize these diverse and independent models by stacking another machine learning algorithm on top of these predictions.…”

Section: ) Stackingmentioning

confidence: 99%

“…Furthermore, medical imaging datasets are commonly quite small, which is why ensemble learning techniques for efficient training data usage are especially popular as demonstrated in Bibaut et al [14] and Müller et al [20]. Empirically, ensemble learning based pipelines tend to be superior according to the assumption that the assembling of diverse models has the advantage to combine their strengths in focusing on different features whereas balancing out the individual incapability of a model [14], [21]- [23]. However, it is still an open question to what extent as well as which ensemble learning strategies are beneficial in deep learning based MIC pipelines.…”

Section: Introductionmentioning

confidence: 99%

“…Even so, the field and idea of general ensemble learning is not novel, the impact of ensemble learning strategies in deep learning based classification has not been adequately analyzed in the literature, yet. Whereas multiple authors provide extensive reviews on general ensemble learning like Ganaiea et al [23], only a handful of works started to survey the deep ensemble learning field. While Cao et al reviewed deep learning based ensemble learning methods specifically in bioinformatics [24], Sagi et al [25], Ju et al [14], and Kandel et al [26] started to provide descriptions or analysis on general deep ensemble learning methods.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

An Analysis on Ensemble Learning Optimized Medical Image Classification With Deep Convolutional Neural Networks

Müller

Soto-Rey

2022

IEEE Access

View full text Add to dashboard Cite

Novel and high-performance medical image classification pipelines are heavily utilizing ensemble learning strategies. The idea of ensemble learning is to assemble diverse models or multiple predictions and, thus, boost prediction performance. However, it is still an open question to what extent as well as which ensemble learning strategies are beneficial in deep learning based medical image classification pipelines. In this work, we proposed a reproducible medical image classification pipeline for analyzing the performance impact of the following ensemble learning techniques: Augmenting, Stacking, and Bagging. The pipeline consists of state-of-the-art preprocessing and image augmentation methods as well as 9 deep convolution neural network architectures. It was applied on four popular medical imaging datasets with varying complexity. Furthermore, 12 pooling functions for combining multiple predictions were analyzed, ranging from simple statistical functions like unweighted averaging up to more complex learning-based functions like support vector machines. Our results revealed that Stacking achieved the largest performance gain of up to 13% F1-score increase. Augmenting showed consistent improvement capabilities by up to 4% and is also applicable to single model based pipelines. Cross-validation based Bagging demonstrated significant performance gain close to Stacking, which resulted in an F1-score increase up to +11%. Furthermore, we demonstrated that simple statistical pooling functions are equal or often even better than more complex pooling functions. We concluded that the integration of ensemble learning techniques is a powerful method for any medical image classification pipeline to improve robustness and boost performance.

show abstract

Section: Ensemble Learning Techniquesmentioning

confidence: 99%

Section: Ensemble Learning Techniquesmentioning

confidence: 99%

Section: ) Stackingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

An Analysis on Ensemble Learning Optimized Medical Image Classification With Deep Convolutional Neural Networks

Müller

Soto-Rey

2022

IEEE Access

View full text Add to dashboard Cite

show abstract

A Competition, Benchmark, Code, and Data for Using Artificial Intelligence to Detect Lesions in Digital Breast Tomosynthesis

Konz

Buda

et al. 2023

JAMA Netw Open

View full text Add to dashboard Cite

ImportanceAn accurate and robust artificial intelligence (AI) algorithm for detecting cancer in digital breast tomosynthesis (DBT) could significantly improve detection accuracy and reduce health care costs worldwide.ObjectivesTo make training and evaluation data for the development of AI algorithms for DBT analysis available, to develop well-defined benchmarks, and to create publicly available code for existing methods.Design, Setting, and ParticipantsThis diagnostic study is based on a multi-institutional international grand challenge in which research teams developed algorithms to detect lesions in DBT. A data set of 22 032 reconstructed DBT volumes was made available to research teams. Phase 1, in which teams were provided 700 scans from the training set, 120 from the validation set, and 180 from the test set, took place from December 2020 to January 2021, and phase 2, in which teams were given the full data set, took place from May to July 2021.Main Outcomes and MeasuresThe overall performance was evaluated by mean sensitivity for biopsied lesions using only DBT volumes with biopsied lesions; ties were broken by including all DBT volumes.ResultsA total of 8 teams participated in the challenge. The team with the highest mean sensitivity for biopsied lesions was the NYU B-Team, with 0.957 (95% CI, 0.924-0.984), and the second-place team, ZeDuS, had a mean sensitivity of 0.926 (95% CI, 0.881-0.964). When the results were aggregated, the mean sensitivity for all submitted algorithms was 0.879; for only those who participated in phase 2, it was 0.926.Conclusions and RelevanceIn this diagnostic study, an international competition produced algorithms with high sensitivity for using AI to detect lesions on DBT images. A standardized performance benchmark for the detection task using publicly available clinical imaging data was released, with detailed descriptions and analyses of submitted algorithms accompanied by a public release of their predictions and code for selected methods. These resources will serve as a foundation for future research on computer-assisted diagnosis methods for DBT, significantly lowering the barrier of entry for new researchers.

show abstract

Machine Learning‐Enabled Precision Position Control and Thermal Regulation in Advanced Thermal Actuators

Mirvakili,

Haghighat,

Sim

2024

Adv Materials Technologies

View full text Add to dashboard Cite

With their unique combination of characteristics – an energy density almost 100 times that of human muscle, and a power density of 5.3 kW kg−1, similar to a jet engine's output – Nylon artificial muscles stand out as particularly apt for robotics applications. However, the necessity of integrating sensors and controllers poses a limitation to their practical usage. Here, a constant power open‐loop controller is reported based on machine learning. It shows that the position of a nylon artificial muscle without external sensors can be controlled. To this end, a mapping is constructed from a desired displacement trajectory to a required power using an ensemble encoder‐style feed‐forward neural network. The neural controller is carefully trained on a physics‐based denoised dataset and can be fine‐tuned to accommodate various types of thermal artificial muscles, irrespective of the presence or absence of hysteresis. This neural network effectively endows the artificial muscles with “muscle memory”, allowing them to replicate complex movements reliably.

show abstract

Ensemble deep learning: A review

Cited by 755 publications

References 181 publications

An Analysis on Ensemble Learning Optimized Medical Image Classification With Deep Convolutional Neural Networks

An Analysis on Ensemble Learning Optimized Medical Image Classification With Deep Convolutional Neural Networks

A Competition, Benchmark, Code, and Data for Using Artificial Intelligence to Detect Lesions in Digital Breast Tomosynthesis

Machine Learning‐Enabled Precision Position Control and Thermal Regulation in Advanced Thermal Actuators

Contact Info

Product

Resources

About