arXiv research

Detects backdoors in trained models without poisoned training data.

The paper analyzes the maximum margin algorithm's performance on noisy data.

A new framework selects information sources to test hypotheses robustly, even with misclassifications.

Optimal subset selection for hypothesis testing with penalties.

The study sets limits on how well halfspaces can be learned when labels are corrupted.

Community detection is a fundamental statistical problem in network data analysis. Many algorithms have been proposed to tackle this problem. Most of these algorithms are not guaranteed to achieve the statistical optimality of the problem, while procedures that achieve information theoretic limits for general parameter…

Sparse multinomial logistic regression for multiclass classification with feature selection.

We discuss the problem of risk estimation in the classification problem, with specific focus on finding distributions that maximize the confidence intervals of risk estimation. We derived simple analytic approximations for the maximum bias of empirical risk for histogram classifier. We carry out a detailed study on usi…

We address the problem of learning to benchmark the best achievable classifier performance. In this problem the objective is to establish statistically consistent estimates of the Bayes misclassification error rate without having to learn a Bayes-optimal classifier. Our learning to benchmark framework improves on previ…

NeuroMem-FHP framework estimates FHP parameters efficiently.

New research shows logistic regression can achieve optimal error rate for agnostic learning of halfspaces.

New algorithm detects communities near KS threshold with optimal rate, even in noisy conditions.

Based on the relations between scattering operators of asymptotically hyperbolic metrics and Dirichlet-to-Neumann operators of uniformly degenerate elliptic boundary value problems, we formulate fractional Yamabe problems that include the boundary Yamabe problem studied by Escobar. We observe an interesting Hopf type m…

New attacks exploit transfer learning to misclassify text models.

Study how noisy labels affect semi-supervised learning.

Deep neural networks (DNNs) have achieved exceptional performances in many tasks, particularly, in supervised classification tasks. However, achievements with supervised classification tasks are based on large datasets with well-separated classes. Typically, real-world applications involve wild datasets that include si…

Study identifies and mitigates causes of image misclassifications in CNN models.

Proposes a novel SVM model for binary classification with different misclassification costs.

We describe a new interpretation of the fractional GJMS operators as generalized Dirichlet-to-Neumann operators associated to weighted GJMS operators on naturally associated smooth metric measure spaces. This gives a geometric interpretation of the Caffarelli--Silvestre extension for $(-Δ)^γ$ when $γ\in(0,1)$, and both…

A new method for high-dimensional data classification reduces misclassification errors.

BOBYQA optimizes deep networks with fewer queries than other methods.

SEDA improves RLDA for high-dimensional data.

Dynamics and function of neuronal networks are determined by their synaptic connectivity. Current experimental methods to analyze synaptic network structure on the cellular level, however, cover only small fractions of functional neuronal circuits, typically without a simultaneous record of neuronal spiking activity. H…

The paper improves uncertainty quantification for node classification using distance-based regularization.

The paper derives formulas for option pricing and random walk expectations.

Paper introduces a new uncertainty measure for misclassification detection.

Recent research has demonstrated the brittleness of machine learning systems to adversarial perturbations. However, the studies have been mostly limited to perturbations on images and more generally, classification that does not deal with temporally varying inputs. In this paper we ask "Are adversarial perturbations po…

Study improves adversarial classification using distributionally robust models.

PANDA improves linear discriminant analysis in high dimensions with minimal tuning.

Stock trend prediction is a challenging task due to the market's noise, and machine learning techniques have recently been successful in coping with this challenge. In this research, we create a novel framework for stock prediction, Dynamic Advisor-Based Ensemble (dynABE). dynABE explores domain-specific areas based on…

New method for multiclass classification reduces error bounds.

Non-convex SGD learns halfspaces with adversarial label noise efficiently.

Method minimizes total cost of classification by acquiring covariates efficiently.

Polynomial-time tester-learner for general halfspaces with Gaussian adversarial noise.

This paper proposes CSADA to make DNNs cost-sensitive.

The paper explores how benign overfitting occurs in heavy-tailed input distributions.

Algorithm learns halfspaces in noisy data efficiently.

A fast method computes class-specific adversarial perturbations for deep networks.

This work further develops the properties of fractional differential forms. In particular, finite dimensional subspaces of fractional form spaces are considered. An inner product, Hodge dual, and covariant derivative are defined. Coordinate transformation rules for integral order forms are also computed. Matrix order f…

Efficient algorithm for CLSBM reduces misclassification rate.

Composite likelihood inference of fractional Gaussian processes with sequentially optimal subset selection

Statistical uncertainty of different filtration techniques for market network analysis is studied. Two measures of statistical uncertainty are discussed. One is based on conditional risk for multiple decision statistical procedures and another one is based on average fraction of errors. It is shown that for some import…

We analyze the problem of learning a single user's preferences in an active learning setting, sequentially and adaptively querying the user over a finite time horizon. Learning is conducted via choice-based queries, where the user selects her preferred option among a small subset of offered alternatives. These queries …

With the growth of adversarial attacks against machine learning models, several concerns have emerged about potential vulnerabilities in designing deep neural network-based intrusion detection systems (IDS). In this paper, we study the resilience of deep learning-based intrusion detection systems against adversarial at…

New scoring rules improve probabilistic classification model evaluation.

FPG uses fractional calculus for efficient reinforcement learning with long-term memory.

We consider high-dimensional binary classification by sparse logistic regression. We propose a model/feature selection procedure based on penalized maximum likelihood with a complexity penalty on the model size and derive the non-asymptotic bounds for the resulting misclassification excess risk. The bounds can be reduc…

We carefully study how well minimizing convex surrogate loss functions, corresponds to minimizing the misclassification error rate for the problem of binary classification with linear predictors. In particular, we show that amongst all convex surrogate losses, the hinge loss gives essentially the best possible bound, o…