A5014674732- Adversarial Robustness in Machine Learning
- Machine Learning and Algorithms
- Imbalanced Data Classification Techniques
- Anomaly Detection Techniques and Applications
- Fault Detection and Control Systems
- Statistical Methods and Inference
- Computational Physics and Python Applications
- Security in Wireless Sensor Networks
- Constraint Satisfaction and Optimization
- Image and Signal Denoising Methods
- Ethics and Social Impacts of AI
- Reinforcement Learning in Robotics
- Machine Learning and Data Classification
- Logic, Reasoning, and Knowledge
- Optimization and Search Problems
- Neural Networks and Applications
- Control Systems and Identification
- Model Reduction and Neural Networks
- Advanced Vision and Imaging
- Pediatric Hepatobiliary Diseases and Treatments
- Text and Document Classification Technologies
- Video Analysis and Summarization
- Energy Harvesting in Wireless Networks
- Machine Learning in Bioinformatics
- Distributed Sensor Networks and Detection Algorithms
Dali University
2024
Courant Institute of Mathematical Sciences
2020-2024
New York University
2020-2021
Class imbalance remains a major challenge in machine learning, especially multi-class problems with long-tailed distributions. Existing methods, such as data resampling, cost-sensitive techniques, and logistic loss modifications, though popular often effective, lack solid theoretical foundations. As an example, we demonstrate that methods are not Bayes consistent. This paper introduces novel framework for analyzing generalization imbalanced classification. We propose new class-imbalanced...
Abstract We discuss two key problems related to learning and optimization of neural networks: the computation adversarial attack for robustness approximate complex functions. show that both can be cast as instances DC-programming. give an explicit decomposition corresponding functions differences convex (DC) report results experiments demonstrating effectiveness DCA algorithm applied these problems.
Adversarial robustness is an increasingly critical property of classifiers in applications. The design robust algorithms relies on surrogate losses since the optimization adversarial loss with most hypothesis sets NP-hard. But which should be used and when do they benefit from theoretical guarantees? We present extensive study this question, including a detailed analysis H-calibration H-consistency losses. show that, under some general assumptions, convex functions, or supremum-based often...
We present a detailed study of $H$-consistency bounds for regression. first new theorems that generalize the tools previously given to establish bounds. This generalization proves essential analyzing specific Next, we prove series novel surrogate loss functions squared loss, under assumption symmetric distribution and bounded hypothesis set. includes positive results Huber all $\ell_p$ losses, $p \geq 1$, $\epsilon$-insensitive as well negative result used in Support Vector Regression (SVR)....
This paper presents a comprehensive analysis of the growth rate $H$-consistency bounds (and excess error bounds) for various surrogate losses used in classification. We prove square-root near zero smooth margin-based binary classification, providing both upper and lower under mild assumptions. result also translates to bounds. Our bound requires weaker conditions than those previous work bounds, our is entirely novel. Moreover, we extend this multi-class classification with series novel...
Recent research has introduced a key notion of $H$-consistency bounds for surrogate losses. These offer finite-sample guarantees, quantifying the relationship between zero-one estimation error (or other target loss) and loss specific hypothesis set. However, previous were derived under condition that lower bound conditional regret is given as convex function regret, without non-constant factors depending on predictor or input instance. Can we derive finer more favorable bounds? In this work,...
By combining unsupervised and supervised machine learning methods, we have proposed a framework, called USmorph , to carry out automatic classifications of galaxy morphologies. In this work, update the (UML) step by proposing an algorithm based on ConvNeXt large model coding improve efficiency unlabeled morphology classifications. The method can be summarized into three key aspects as follows: (1) convolutional autoencoder is used for image denoising reconstruction rotational invariance...
We present a detailed study of surrogate losses and algorithms for multi-label learning, supported by $H$-consistency bounds. first show that, the simplest form loss (the popular Hamming loss), well-known consistent binary relevance suffers from sub-optimal dependency on number labels in terms bounds, when using smooth such as logistic losses. Furthermore, this function fails to account label correlations. To address these drawbacks, we introduce novel loss, that accounts correlations...
We present a comprehensive study of surrogate loss functions for learning to defer. introduce broad family losses, parameterized by non-increasing function $\Psi$, and establish their realizable $H$-consistency under mild conditions. For cost based on classification error, we further show that these losses admit bounds when the hypothesis set is symmetric complete, property satisfied common neural network linear sets. Our results also resolve an open question raised in previous work...
We present a more general analysis of $H$-calibration for adversarially robust classification. By adopting finer definition calibration, we can cover settings beyond the restricted hypothesis sets studied in previous work. In particular, our results hold most common used machine learning. both fix some calibration (Bao et al., 2020) and generalize others (Awasthi 2021). Moreover, results, combined with study consistency by Awasthi al. (2021), also lead to $H$-consistency covering sets.