Test-time adversarial attacks have posed serious challenges to the robustness of machine-learning models, and in many settings the adversarial perturbation need not be bounded by small $\ell_p$-norms. Motivated by the semantics-preserving attacks in vision and security domain, we investigate $\textit{relational adversaries}$, a broad class of attackers who create adversarial examples that are in a reflexive-transitive closure of a logical relation. We analyze the conditions for robustness and propose $\textit{normalize-and-predict}$ -- a learning framework with provable robustness guarantee. We compare our approach with adversarial training and derive an unified framework that provides benefits of both approaches. Guided by our theoretical findings, we apply our framework to image classification and malware detection. Results of both tasks show that attacks using relational adversaries frequently fool existing models, but our unified framework can significantly enhance their robustness.

Probability is the cornerstone of Artificial Intelligence. The management of uncertainty is key to many applications of AI, such as machine learning, filtering, robotics, computer vision, NLP, search and so on. And no other sector is the management of uncertainty as crucial as it is in the health sector. At first glance, the false-negative seems more devastating. Of course, a false allergy test-result has the likely outcome of a GP administering a drug to you that could cause life-threatening issues.

The environments in which we deploy machine learning (ML) algorithms rarely look exactly like the environments in which we collected our training data. Unfortunately, we lack methodology for evaluating how well an algorithm will generalize to new environments that differ in a structured way from the training data (i.e., the case of dataset shift (Quiñonero-Candela et al., 2009)). Such methodology is increasingly important as ML systems are being deployed across a number of industries, such as health care and personal finance, in which system performance translates directly to real-world outcomes. Further, as regulation and product reviews become more common across industries, system developers will be expected to produce evidence of the validity and safety of their systems. For example, the United States Food and Drug Administration (FDA) currently regulates ML systems for medical applications, requiring evidence for the validity of such systems before approval is granted (US Food and Drug Administration, 2019). Evaluation methods for assessing model validity have typically focused on how the model performs on data from the training distribution, known as internal validity. Powerful tools, such as cross-validation and the bootstrap, satisfy the assumption that the training and test data are drawn from the same distribution. However, these validation methods do not capture a model's ability to generalize to new environments, known as external validity (Campbell and Stanley, 1963). Currently, the main way to assess a model's external validity is to empirically evaluate performance on multiple, independently collected datasets (e.g.,

Diabetes is one of the most common disease in individuals. \textit{Diabetic retinopathy} (DR) is a complication of diabetes, which could lead to blindness. Automatic DR grading based on retinal images provides a great diagnostic and prognostic value for treatment planning. However, the subtle differences among severity levels make it difficult to capture important features using conventional methods. To alleviate the problems, a new deep learning architecture for robust DR grading is proposed, referred to as SEA-Net, in which, spatial attention and channel attention are alternatively carried out and boosted with each other, improving the classification performance. In addition, a hybrid loss function is proposed to further maximize the inter-class distance and reduce the intra-class variability. Experimental results have shown the effectiveness of the proposed architecture.

Multiclass classifiers are often designed and evaluated only on a sample from the classes on which they will eventually be applied. Hence, their final accuracy remains unknown. In this work we study how a classifier's performance over the initial class sample can be used to extrapolate its expected accuracy on a larger, unobserved set of classes. For this, we define a measure of separation between correct and incorrect classes that is independent of the number of classes: the reversed ROC (rROC), which is obtained by replacing the roles of classes and data-points in the common ROC. We show that the classification accuracy is a function of the rROC in multiclass classifiers, for which the learned representation of data from the initial class sample remains unchanged when new classes are added. Using these results we formulate a robust neural-network-based algorithm, CleaneX, which learns to estimate the accuracy of such classifiers on arbitrarily large sets of classes. Our method achieves remarkably better predictions than current state-of-the-art methods on both simulations and real datasets of object detection, face recognition, and brain decoding.

Being to new text analytics, I haven't gotten the hang of my typical ML workflow given how long processes take to run in the commonly large feature space of text analytics. I would like to know what the typical strategy is to balance effort/time in terms of optimizing transformation decision, feature down-selection, and model tuning. In an effort to get a sense of which of the decision points above I should run further tuning on, I ran untuned RF, Logistic, Naive Bayes, SGD, and KNN models on (with cross validation). No clear decision point was commonly "better" in the resulting f-1 scores, and the difference is often noteworthy. As I have no bias towards a particular algorithm type (only the best f-1 score), I'm stuck in a quandry-- I have not successfully narrowed my decision space enough.

This article is to set up the framework with a simple model with a detailed walk through of each step. There are tons of improvements that can be made to boost model performance! In the world of healthcare, one of the major issues that medical professionals face is the correct diagnosis of conditions and diseases of patients. Not being able to correctly diagnose a condition is a problem for both the patient and the doctor. The doctor is not benefiting the patient in the appropriate way if the doctor misdiagnoses the patient.

Selective classification, in which models are allowed to abstain on uncertain predictions, is a natural approach to improving accuracy in settings where errors are costly but abstentions are manageable. In this paper, we find that while selective classification can improve average accuracies, it can simultaneously magnify existing accuracy disparities between various groups within a population, especially in the presence of spurious correlations. We observe this behavior consistently across five datasets from computer vision and NLP. Surprisingly, increasing the abstention rate can even decrease accuracies on some groups. To better understand when selective classification improves or worsens accuracy on a group, we study its margin distribution, which captures the model's confidences over all predictions. For example, when the margin distribution is symmetric, we prove that whether selective classification monotonically improves or worsens accuracy is fully determined by the accuracy at full coverage (i.e., without any abstentions) and whether the distribution satisfies a property we term left-log-concavity. Our analysis also shows that selective classification tends to magnify accuracy disparities that are present at full coverage. Fortunately, we find that it uniformly improves each group when applied to distributionally-robust models that achieve similar full-coverage accuracies across groups. Altogether, our results imply selective classification should be used with care and underscore the importance of models that perform equally well across groups at full coverage.

The Standard Model of particle physics (SM) [1, 2] is a remarkably effective theory, able to describe the experimental observations made thus far in high energy physics with unprecedented precision and completeness. Despite its success however, this model fails to explain several observations like the baryon asymmetry and the presence of dark matter, which inspires searches for extensions to the SM. The study of the recently discovered [3-5] Higgs boson [6-11], and the search for extensions of the electroweak sector are two of the most active research sectors in the field. Because of the flavour asymmetry associated to production and decay processes in each case, the ability to classify jets originating from heavy-flavour (bottom and charm) quarks is important. Heavy-flavour (HF) jets contain an open-bottom or open-charm hadron as a result of the fragmentation process. This hadron carries a large fraction of the initial parton momentum. HF hadrons also have a sizeable lifetime, with a of 0.5 mm and 0.3 mm for bottom and charm, respectively.

Unlike in the traditional statistical modeling for which a user typically hand-specify a prior, Neural Processes (NPs) implicitly define a broad class of stochastic processes with neural networks. Given a data stream, NP learns a stochastic process that best describes the data. While this "data-driven" way of learning stochastic processes has proven to handle various types of data, NPs still rely on an assumption that uncertainty in stochastic processes is modeled by a single latent variable, which potentially limits the flexibility. To this end, we propose the Boostrapping Neural Process (BNP), a novel extension of the NP family using the bootstrap. The bootstrap is a classical data-driven technique for estimating uncertainty, which allows BNP to learn the stochasticity in NPs without assuming a particular form. We demonstrate the efficacy of BNP on various types of data and its robustness in the presence of model-data mismatch.