arXiv:2505.24452v4 Announce Type: replace 
Abstract: The expanding computational costs and limited resources underscore the critical need for budgeted-iteration training, which aims to achieve optimal learning within predetermined iteration budgets. While learning rate schedules fundamentally govern the performance of different networks and tasks, particularly in budgeted-iteration scenarios, their design remains largely heuristic, lacking theoretical foundations. In addition, the optimal learning rate schedule requires extensive trial-and-error selection, making the training process inefficient. In this work, we propose the Unified Budget-Aware (UBA) schedule, a theoretically grounded learning rate schedule that consistently outperforms commonly-used schedules among diverse architectures and tasks under different constrained training budgets. First, we bridge the gap by constructing a novel training budget-aware optimization framework, which explicitly accounts for the robustness to landscape curvature variations. From this framework, we derive the UBA schedule, controlled by a single hyper-parameter \varphi that provides a trade-off between flexibility and simplicity, eliminating the need for per-network numerical optimization. Moreover, we establish a theoretical connection between \varphi and the condition number, adding interpretation and justification to our approach. Besides, we prove the convergence for different values of \varphi. We offer practical guidelines for its selection via theoretical analysis and empirical results. Extensive experimental results show that UBA consistently surpasses the commonly-used schedules across diverse vision and language tasks, spanning network architectures (e.g., ResNet, OLMo) and scales, under different training-iteration budgets.

تقدم دراسة جديدة جدول زمني موحد لمعدل التعلم واعي للميزانية (UBA)، مصمم لتحسين التدريب ضمن ميزانيات تكرارية مقيدة. يتناول هذا النهج عدم الكفاءة في الجداول الزمنية لمعدل التعلم التقليدية، التي تعتمد غالبًا على طرق تجريبية وواسعة النطاق. لقد أظهر جدول UBA أداءً متفوقًا عبر هياكل ومهام متنوعة، خاصة في سيناريوهات التدريب ذات الميزانية المحدودة.

Un nuevo estudio presenta el calendario de tasa de aprendizaje unificado y consciente del presupuesto (UBA), diseñado para optimizar el entrenamiento dentro de presupuestos de iteración restringidos. Este enfoque aborda las ineficiencias de los calendarios de tasa de aprendizaje tradicionales, que a menudo dependen de métodos heurísticos y de una extensa prueba y error. El calendario UBA ha demostrado un rendimiento superior en diversas arquitecturas y tareas, particularmente en escenarios de iteración presupuestada.

Une nouvelle étude présente le calendrier d'apprentissage unifié et conscient du budget (UBA), conçu pour optimiser l'entraînement dans des budgets d'itération contraints. Cette approche répond aux inefficacités des calendriers d'apprentissage traditionnels, qui reposent souvent sur des méthodes heuristiques et des essais-erreurs étendus. Le calendrier UBA a montré une performance supérieure à travers diverses architectures et tâches, en particulier dans des scénarios d'itération budgétisée.

A new study introduces the Unified Budget-Aware (UBA) learning rate schedule, designed to optimize training within constrained iteration budgets. This approach addresses the inefficiencies of traditional learning rate schedules, which often rely on heuristic methods and extensive trial-and-error. The UBA schedule has shown superior performance across various architectures and tasks, particularly in budgeted-iteration scenarios.

Stepsize anything: A unified learning rate schedule for budgeted-iteration training

arXiv:2512.06377v1 Announce Type: new 
Abstract: Current FER (Facial Expression Recognition) dataset is mostly labeled by emotion categories, such as happy, angry, sad, fear, disgust, surprise, and neutral which are limited in expressiveness. However, future affective computing requires more comprehensive and precise emotion metrics which could be measured by VAD(Valence-Arousal-Dominance) multidimension parameters. To address this, AffectNet has tried to add VA (Valence and Arousal) information, but still lacks D(Dominance). Thus, the research introduces VAD annotation on FER2013 dataset, takes the initiative to label D(Dominance) dimension. Then, to further improve network capacity, it enforces orthogonalized convolution on it, which extracts more diverse and expressive features and will finally increase the prediction accuracy. Experiment results show that D dimension could be measured but is difficult to obtain compared with V and A dimension no matter in manual annotation or regression network prediction. Secondly, the ablation test by introducing orthogonal convolution verifies that better VAD prediction could be obtained in the configuration of orthogonal convolution. Therefore, the research provides an initiative labelling for D dimension on FER dataset, and proposes a better prediction network for VAD prediction through orthogonal convolution. The newly built VAD annotated FER2013 dataset could act as a benchmark to measure VAD multidimensional emotions, while the orthogonalized regression network based on ResNet could act as the facial expression recognition baseline for VAD emotion prediction. The newly labeled dataset and implementation code is publicly available on https://github.com/YeeHoran/VAD-Net .

تقدم دراسة جديدة VAD-Net، وهو نهج متعدد الأبعاد للتعرف على تعبيرات الوجه (FER) الذي يعزز مجموعة بيانات FER2013 الحالية من خلال دمج البعد D (الهيمنة) جنبًا إلى جنب مع مقاييس القيمة والاهتمام. يهدف هذا التقدم إلى تحسين دقة التعرف على المشاعر في أنظمة التعليم الذكي، مع معالجة قيود طرق تصنيف المشاعر الحالية.

Un nuevo estudio presenta VAD-Net, un enfoque multidimensional para el reconocimiento de expresiones faciales (FER) que mejora el conjunto de datos FER2013 existente al incorporar la dimensión D (Dominancia) junto con las métricas de Valencia y Arousal. Este avance tiene como objetivo mejorar la precisión del reconocimiento emocional en sistemas educativos inteligentes, abordando las limitaciones de los métodos actuales de categorización emocional.

Une nouvelle étude présente VAD-Net, une approche multidimensionnelle de la reconnaissance des expressions faciales (FER) qui améliore le jeu de données FER2013 existant en intégrant la dimension D (Dominance) aux métriques de Valence et d'Arousal. Cette avancée vise à améliorer la précision de la reconnaissance des émotions dans les systèmes éducatifs intelligents, en répondant aux limites des méthodes de catégorisation des émotions actuelles.

A new study introduces VAD-Net, a multidimensional approach to facial expression recognition (FER) that enhances the existing FER2013 dataset by incorporating the D (Dominance) dimension alongside Valence and Arousal metrics. This advancement aims to improve the precision of emotion recognition in intelligent education systems, addressing the limitations of current emotion categorization methods.

VAD-Net: Multidimensional Facial Expression Recognition in Intelligent Education System

arXiv:2410.09101v2 Announce Type: replace-cross 
Abstract: Dataset ownership verification, the process of determining if a dataset is used in a model's training data, is necessary for detecting unauthorized data usage and data contamination. Existing approaches, such as backdoor watermarking, rely on inducing a detectable behavior into the trained model on a part of the data distribution. However, these approaches have limitations, as they can be harmful to the model's performances or require unpractical access to the model's internals. Most importantly, previous approaches lack guarantee against false positives. This paper introduces data taggants, a novel non-backdoor dataset ownership verification technique. Our method uses pairs of out-of-distribution samples and random labels as secret keys, and leverages clean-label targeted data poisoning to subtly alter a dataset, so that models trained on it respond to the key samples with the corresponding key labels. The keys are built as to allow for statistical certificates with black-box access only to the model. We validate our approach through comprehensive and realistic experiments on ImageNet1k using ViT and ResNet models with state-of-the-art training recipes. Our findings demonstrate that data taggants can reliably detect models trained on the protected dataset with high confidence, without compromising validation accuracy, and show their superiority over backdoor watermarking. We demonstrate the stealthiness and robustness of our method against various defense mechanisms.

يقدم مقال جديد تقنية تسمى علامات البيانات، وهي طريقة للتحقق من ملكية مجموعات البيانات تستخدم التسمم المستهدف للبيانات بشكل غير ضار لتعديل مجموعات البيانات بشكل طفيف. تهدف هذه الطريقة إلى معالجة قيود الأساليب الحالية، مثل العلامات الخلفية، التي يمكن أن تضر بأداء النموذج وتفتقر إلى الضمانات ضد الإيجابيات الكاذبة.

Un nuevo artículo presenta los taggants de datos, una técnica para la verificación de la propiedad de conjuntos de datos que utiliza envenenamiento de datos dirigido inofensivo para alterar sutilmente los conjuntos de datos. Este método busca abordar las limitaciones de enfoques existentes, como el marcado de puerta trasera, que pueden perjudicar el rendimiento del modelo y carecen de garantías contra falsos positivos.

Un nouvel article présente les taggants de données, une technique de vérification de la propriété des ensembles de données qui utilise un empoisonnement de données ciblé inoffensif pour altérer subtilement les ensembles de données. Cette méthode vise à surmonter les limitations des approches existantes, telles que le marquage par porte dérobée, qui peuvent nuire aux performances des modèles et manquent de garanties contre les faux positifs.

A new paper introduces data taggants, a technique for dataset ownership verification that utilizes harmless targeted data poisoning to subtly alter datasets. This method aims to address the limitations of existing approaches, such as backdoor watermarking, which can harm model performance and lack guarantees against false positives.

Stepsize anything: A unified learning rate schedule for budgeted-iteration training

Was this article worth reading? Share it

LucidQuery AI

Hypertune

MicroEstimates