Deep MLPs for Imbalanced Classification

Díaz-Vico, David; Figueiras-Vidal, Anı́bal R.; Dorronsoro, José R.

doi:10.1109/ijcnn.2018.8489504

Cited by 12 publications

(10 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…CLEMS (Huang and Lin, 2017) introduces a cost-sensitive label embedding technique that takes the cost function of interest into account. CS-DMLP (Díaz-Vico et al, 2018) is a deep multi-layer percetron model utilizing cost-sensitive learning to regularize the posterior probability distribution predicted for a given sample. This type of methods normally requires domain knowledge to define the actual cost value, which is often hard in real-world scenarios (Krawczyk, 2016).…”

Section: Algorithm-level Methodsmentioning

confidence: 99%

“…The most popular branch is the costsensitive algorithms, which assign a higher cost on misclassifying the minority class instances. (Díaz-Vico et al, 2018). (3) Ensemble-based methods that combine advantages of data-level and algorithmlevel methods by merging data-level solutions with classifier ensembles, resulting in robust and efficient learners (Galar et al, 2012;Wang et al, 2015).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

SetConv: A New Approach for Learning from Imbalanced Data

Gao¹,

Li²,

Yang³

et al. 2020

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)

View full text Add to dashboard Cite

For many real-world classification problems, e.g., sentiment classification, most existing machine learning methods are biased towards the majority class when the Imbalance Ratio (IR) is high. To address this problem, we propose a set convolution (SetConv) operation and an episodic training strategy to extract a single representative for each class, so that classifiers can later be trained on a balanced class distribution. We prove that our proposed algorithm is permutation-invariant despite the order of inputs, and experiments on multiple large-scale benchmark text datasets show the superiority of our proposed framework when compared to other SOTA methods.

show abstract

Section: Algorithm-level Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

SetConv: A New Approach for Learning from Imbalanced Data

Gao¹,

Li²,

Yang³

et al. 2020

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)

View full text Add to dashboard Cite

show abstract

“…SCUT is a hybrid sampling method proposed by Agarwal et al [26] for balancing the examples in multiclass data sets. The minority classes are oversampled generating synthetic examples with SMOTE while ( ) SCUT, SMOTE and cluster-based undersampling [26] MDO, Mahalanobis based oversampling [27,28,29] ( ) ( ) SMOM, synthetic oversampling for multiclass [30] ( ) ( ) Hellinger distance decision trees [31] ( ) Dynamic sampling for multilayer perceptrons [32] Deep MLPs for imbalance [33] AdaC2.M1, cost-sensitive boosting [34] Cost sensitive OVO ensemble [35] Cost-Sensitive neural networks with binarization [36] OVA with hybrid sampling [9] OVO fuzzy rough set [37] Binarization with over/undersampling [38] ( ) ( ) ( ) ( ) Instance weighting (cost-sensitive) [38] ( ) ( ) UnderBagging [39,17] ( ) SMOTEBagging [40,17] ( ) RUSBoost [41,17] ( ) SMOTEBoost [42,17] ( ) SMOTE+AdaBoost [17] ( ) EasyEnsemble [43,17,44] ( ) Binarization with boosting and oversampling [45] Diversified ECOC [46] RAMOBoost [47,32] AdaBoost.NC [48,17,38,44] ( ) ( ) ( ) ( ) Probability threshold Bagging [49] Dynamic ensemble selection [44] Multiclass Roughly Balanced Bagging [50,49] ( ) ( ) the m...…”

Section: Multiclass Imbalanced Classificationmentioning

confidence: 99%

“…Sáez et al [22] found that oversampling benefits from distinguishing between four example Algorithm level. Publications reporting methods adapted to multiclass imbalance are [31], [32] and [33].…”

Section: Multiclass Imbalanced Classificationmentioning

confidence: 99%

See 1 more Smart Citation

Random Balance ensembles for multiclass imbalance learning

Rodrı́guez

Díez-Pastor

Arnaiz‐González

et al. 2020

Knowledge-Based Systems

View full text Add to dashboard Cite

Random Balance strategy (RandBal) has been recently proposed for constructing classifier ensembles for imbalanced, two-class data sets. In RandBal, each base classifier is trained with a sample of the data with a random class prevalence, independent of the a priori distribution. Hence, for each sample, one of the classes will be undersampled while the other will be oversampled. RandBal can be applied on its own or can be combined with any other ensemble method. One particularly successful variant is RandBalBoost which integrates Random Balance and boosting. Encouraged by the success of RandBal, this work proposes two approaches which extend RandBal to multiclass imbalance problems. Multiclass imbalance implies that at least two classes have substantially different proportion of instances. In the first approach proposed here, termed Multiple Random Balance (MultiRandBal), we deal with all classes simultaneously. The training data for each base classifier are sampled with random class proportions. The second approach we propose decomposes the multiclass problem into two-class problems using one-vs-one or one-vs-all, and builds an ensemble of RandBal ensembles. We call the two versions of the second approach OVO-RandBal and OVA-RandBal, respectively. These two approaches were chosen because they are the most straightforward extensions of RandBal for multiple classes. Our main objective is to evaluate both approaches for multiclass imbalanced problems. To this end, an experiment was carried out with 52 multiclass data sets. The results suggest that both MultiRandBal, and OVO/OVA-RandBal are viable extensions of the original twoclass RandBal. Collectively, they consistently outperform acclaimed state-of-the art methods for multiclass imbalanced problems.

show abstract

Balancing Agents for Mining Imbalanced Multiclass Datasets – Performance Evaluation

Jȩdrzejowicz

Jędrzejowicz

2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Deep MLPs for Imbalanced Classification

Cited by 12 publications

References 18 publications

SetConv: A New Approach for Learning from Imbalanced Data

SetConv: A New Approach for Learning from Imbalanced Data

Random Balance ensembles for multiclass imbalance learning

Balancing Agents for Mining Imbalanced Multiclass Datasets – Performance Evaluation

Contact Info

Product

Resources

About