A unifying view of class overlap and imbalance: Key concepts, multi-view panorama, and open avenues for research

The combination of class imbalance and overlap is currently one of the most challenging issues in machine learning. While seminal work focused on establishing class overlap as a complicating factor for classification tasks in imbalanced domains, ongoing research mostly concerns the study of their synergy over real-word applications. However, given the lack of a well-formulated definition and measurement of class overlap in real-world domains, especially in the presence of class imbalance, the research community has not yet reached a consensus on the characterisation of both problems. This naturally complicates the evaluation of existing approaches to address these issues simultaneously and prevents future research from moving towards the devise of specialised solutions. In this work, we advocate for a unified view of the problem of class overlap in imbalanced domains. Acknowledging class overlap as the overarching problem - since it has proven to be more harmful for classification tasks than class imbalance - we start by discussing the key concepts associated to its definition, identification, and measurement in real-world domains, while advocating for a characterisation of the problem that attends to multiple sources of complexity. We then provide an overview of existing data complexity measures and establish the link to what specific types of class overlap problems these measures cover, proposing a novel taxonomy of class overlap complexity measures. Additionally, we characterise the relationship between measures, the insights they provide, and discuss to what extent they account for class imbalance. Finally, we systematise the current body of knowledge on the topic across several branches of Machine Learning (Data Analysis, Data Preprocessing, Algorithm Design, and Meta-learning), identifying existing limitations and discussing possible lines for future research.

Explainable image analysis for decision support in medical healthcare

Recent advances in medical imaging and deep learning have enabled the efficient analysis of large databases of images. Notable examples include the analysis of computed tomography (CT), magnetic resonance imaging (MRI), and X-ray. While the automatic classification of images has proven successful, adopting such a paradigm in the medical healthcare setting is unfeasible. Indeed, the physician in charge of the detailed medical assessment and diagnosis of patients cannot trust a deep learning model's decisions without further explanations or insights about their classification outcome. In this study, rather than relying on classification, we propose a new method that leverages deep neural networks to extract a representation of images and further analyze them through clustering, dimensionality reduction for visualization, and class activation mapping. Thus, the system does not make decisions on behalf of physicians. Instead, it helps them make a diagnosis. Experimental results on lung images affected by Pneumonia and Covid-19 lesions show the potential of our method as a tool for decision support in a medical setting. It allows the physician to identify groups of similar images and highlight regions of the input that the model deemed important for its predictions.

The Case for Latent Variable Vs Deep Learning Methods in Misinformation Detection: An Application to COVID-19

The detection and removal of misinformation from social media during high impact events, e.g., COVID-19 pandemic, is a sensitive application since the agency in charge of this process must ensure that no unwarranted actions are taken. This suggests that any automated system used for this process must display both high prediction accuracy as well as high explainability. Although Deep Learning methods have shown remarkable prediction accuracy, accessing the contextual information that Deep Learning-based representations carry is a significant challenge. In this paper, we propose a data-driven solution that is based on a popular latent variable model called Independent Component Analysis (ICA), where a slight loss in accuracy with respect to a BERT model is compensated by interpretable contextual representations. Our proposed solution provides direct interpretability without affecting the computational complexity of the model and without designing a separate system. We carry this study on a novel labeled COVID-19 Twitter dataset that is based on socio-linguistic criteria and show that our model’s explanations highly correlate with humans’ reasoning. © 2021, Springer Nature Switzerland AG.

A Semi-supervised Framework for Misinformation Detection

The spread of misinformation in social media outlets has become a prevalent societal problem and is the cause of many kinds of social unrest. Curtailing its prevalence is of great importance and machine learning has shown significant promise. However, there are two main challenges when applying machine learning to this problem. First, while much too prevalent in one respect, misinformation, actually, represents only a minor proportion of all the postings seen on social media. Second, labeling the massive amount of data necessary to train a useful classifier becomes impractical. Considering these challenges, we propose a simple semi-supervised learning framework in order to deal with extreme class imbalances that has the advantage, over other approaches, of using actual rather than simulated data to inflate the minority class. We tested our framework on two sets of Covid-related Twitter data and obtained significant improvement in F1-measure on extremely imbalanced scenarios, as compared to simple classical and deep-learning data generation methods such as SMOTE, ADASYN, or GAN-based data generation. © 2021, Springer Nature Switzerland AG.