Cost-Sensitive Learning of Deep Feature Representations From Imbalanced Data

By Salman Hameed Khan, Munawar Hayat, Mohammed Bennamoun, Ferdous Sohel, Roberto B. Togneri. 2015

977 citations recorded by the source corpus.

Class imbalance is a common problem in the case of real-world object detection and classification tasks. Data of some classes are abundant, making them an overrepresented majority, and data of other classes are scarce, making them an underrepresented minority. This imbalance makes it challenging for a classifier to appropriately learn the discriminating boundaries of the majority and minority classes. In this paper, we propose a cost-sensitive (CoSen) deep neural network, which can automatically learn robust feature representations for both the majority and minority classes. During training, our learning procedure jointly optimizes the class-dependent costs and the neural network parameters. The proposed approach is applicable to both binary and multiclass problems without any modification. Moreover, as opposed to data-level approaches, we do not alter the original data distribution, which results in a lower computational cost during the training process. We report the results of our experiments on six major image classification data sets and show that the proposed approach significantly outperforms the baseline algorithms. Comparisons with popular data sampling techniques and CoSen classifiers demonstrate the superior performance of our proposed method.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging

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

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Decoupling Representation and Classifier for Long-Tailed Recognition (2019, background): “The loss can vary at class-level for matching a given data distribution and improving the generalization of tail classes (Cui et al., 2019; Khan et al., 2017; Cao et al., 2019; Khan et al., 2019; Huang et al., 2019).”
Class-Balanced Loss Based on Effective Number of Samples (2019, background): “Most of previous efforts on long-tailed imbalanced data can be divided into two regimes: re-sampling [35, 12, 4, 50] (including over-sampling and under-sampling) and cost-sensitive learning [38, 49, 17, 22, 34].”
A systematic study of the class imbalance problem in convolutional neural networks (2017, neutral): “Some previous studies showed results on cost sensitive learning of deep neural networks [24, 25, 26].”
Learning to Reweight Examples for Robust Deep Learning (2018, background): “, 2017), cost-sensitive weighting (Ting, 2000; Khan et al., 2015), and structured margin based objectives (Huang et al.”
SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-year Anniversary (2018, background): “In the 1990s as more data and applications of machine learning and data mining started to become prevalent, an important challenge emerged: how to achieve desired classification accuracy when dealing with data that had significantly skewed class distributions (Sun et al., 2009; He & Garcia, 2009; Ló”
Striking the Right Balance With Uncertainty (2019, methodology): “Compared to a recent costsensitive CNN approach [25], we obtain an impressive absolute performance gain of 15.”
DeepSMOTE: Fusing Deep Learning and SMOTE for Imbalanced Data (2021, background): “sensitive solutions [31], [32], [33] and one-class classification [34], [35], [36] can also be considered as a form of algorithmlevel approaches.”
Learning to Reweight with Deep Interactions (2021, background): “…weighting is to increase the robustness of training, including addressing the problem of imbalanced data (Sun et al. 2007; Dong, Gong, and Zhu 2017; Khan et al. 2018), biased data (Zadrozny 2004; Ren et al. 2018), noisy data (Angluin and Laird 1988; Reed et al. 2014; Sukhbaatar and Fergus 2014;…”
A Systematic Review on Imbalanced Data Challenges in Machine Learning (2019, background): “Cost-sensitive neural network for handling imbalanced data is proposed in Khan et al. (2017).”
Learning Deep Representation for Imbalanced Classification (2016, background): “[22] further seek for joint optimization of the class-sensitive costs and deep features.”
Class Rectification Hard Mining for Imbalanced Deep Learning (2017, background): “However, deep learning is likely to suffer even more from imbalanced data distribution [51, 24, 25, 21] and deep learning of imbalanced data is currently under-studied.”
Contrastive Learning based Hybrid Networks for Long-Tailed Image Classification (2021, positive): “Loss re-weighting is one of the simple but effective ways to tailor the loss function for imbalanced classification, where the basic idea is to upweight the tailed samples and downweight the head samples in the loss function [17].”
Brain MRI analysis for Alzheimer’s disease diagnosis using an ensemble system of deep convolutional neural networks (2018, methodology): “If we define the number of output classes as m, then we get (6) To handle the imbalance in the dataset, we used cost-sensitive training [85].”
The Devil is in Classification: A Simple Framework for Long-tail Instance Segmentation (2020, methodology): “Loss Re-weighting [11], [19], [26], [38], [39] This line of works alleviate the bias by applying diﬀerent weights to diﬀerent samples or categories, such that tail classes or samples receive higher attention during training, thus improving the classiﬁcation performance.”
Distribution Alignment: A Unified Framework for Long-tail Visual Recognition (2021, positive): “Class-level methods typically re-weight the standard loss with category-specific coefficients correlated with the sample distributions [18, 9, 3, 21, 20, 34].”
A Survey on Curriculum Learning (2021, neutral): “The biases include class imbalance and label noise, both of which have been studied as typical problems of machine learning with various practical methods (e.g., [9], [47] for the former and [23], [62], [78], [90] Learner at step 𝑖 in seq.”
Dynamically Weighted Balanced Loss: Class Imbalanced Learning and Confidence Calibration of Deep Neural Networks (2021, neutral): “[33] proposed a cost-sensitive approach where”
CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning (2021, positive): “Prominent works include resampling [7, 3, 4, 14] and re-weighting [21, 9, 6, 41] which re-balance the contribution of each class, while others focus on re-weighting each instance [27, 38, 35, 19].”