The collection of large-scale datasets inevitably introduces noisy labels, leading to a substantial degradation in the performance deep neural networks (DNNs). Although sample selection is mainstream method field learning with which aims mitigate impact labels during model training, testing these methods exhibits significant fluctuations across different noise rates and types. In this paper, we propose Cross-to-Merge Training (C2MT), novel framework that insensitive prior information...

While collecting training data, even with the manual verification of experts from crowdsourcing platforms, eliminating incorrect annotations (noisy labels) completely is difficult and expensive. In dealing datasets that contain noisy labels, over-parameterized deep neural networks (DNNs) tend to overfit, leading poor generalization classification performance. As a result, label learning (NLL) has received significant attention in recent years. Existing research shows although DNNs eventually...

Deep neural networks, as applied in the field of nuclear power fault diagnosis, have garnered significant attention alongside advancements artificial intelligence technology. However, "black box" nature deep learning models has raised concerns regarding their deployment scenarios demanding high safety standards, such plants. In this paper, we innovatively propose utilization an explainable method, grounded game theory, to conduct a detailed analysis diagnostic behavior network By leveraging...

Recent advancements in event-based recognition have demonstrated significant promise, yet most existing approaches rely on extensive training, limiting their adaptability for efficient processing of event-driven visual content. Meanwhile, large language models (LLMs) exhibited remarkable zero-shot capabilities across diverse domains, but application to remains largely unexplored. To bridge this gap, we propose \textbf{LLM-EvGen}, an event representation generator that produces LLM-compatible...

ABSTRACT Background The current focus of cardiac rehabilitation is on adults, with no standard nursing plan available for children congenital heart disease. Therefore, it very necessary to develop a standardized early model disease promote the recovery bodily functions and improve quality life in this population. Purpose This study was designed explore feasibility an graded postoperative evaluate its clinical effect standardization care. Methods One hundred sixteen treated at medical...

Neural Architecture Search (NAS) is an important yet challenging task in network design due to its high computational consumption. To address this issue, we propose the Reinforced Evolutionary (RE- NAS), which evolutionary method with reinforced mutation for NAS. Our integrates into evolution algorithm neural architecture exploration, a controller introduced learn effects of slight modifications and make actions. The guides model population evolve efficiently. Furthermore, as child models...

Online language education has grown in popularity over the past decade or so and gained unprecedented momentum during COVID-19. Against this backdrop, professionals have come under inc...

The daily occurrence of traffic accidents has led to the development 3D reconstruction as a key tool for reconstruction, investigation, and insurance claims. This study proposes novel virtual-real-fusion simulation framework that integrates accident generation, unmanned aerial vehicle (UAV)-based image collection, pipeline with advanced computer vision techniques unsupervised point cloud clustering algorithms. Specifically, micro-traffic simulator an autonomous driving are co-simulated...

We present AutoPose, a novel neural architecture search(NAS) framework that is capable of automatically discovering multiple parallel branches cross-scale connections towards accurate and high-resolution 2D human pose estimation. Recently, high-performance hand-crafted convolutional networks for estimation show growing demands on multi-scale fusion representations. However, current NAS works exhibit limited flexibility scale searching, they dominantly adopt simplified search spaces...

Currently, tourists tend to plan travel routes and itineraries by searching for relevant information on tourist attractions via the Internet intelligent terminals. However, it is difficult achieve good retrieval effect attraction images with text labels. Based deep learning, visual location identification faces such defects as frequent mismatching, high probability of weak matching, long execution time. To solve these defects, this paper puts forward a novel method personalized...

According to security demands of video conference system, this paper presented three schemes encrypt parts data using permutation code and DES encryption algorithm based on the newest coding standard H.264. Comparing with effects two algorithms respectively, determined one as a last scheme. This scheme adopts motion vector residuals after transform quantization before entropy encoding, some codewords DCT coefficients residual in stream, which includes TrailingOnes signs Levels RunBefore. It...

To deal with the inconsistency of minimum variance (MV) benchmark in evaluating non-Gaussian disturbance systems, this paper proposed a new benchmark, which combined entropy output mean value. For cascade control system, was constructed by analyzing weakness MV and pure Renyi benchmark. In order to estimate more accurate performance unknown an improved estimation distribution algorithm based on criterion is given. It can identify calculate index evaluation Finally, different distributions...

AbstractThis article reports on the results of an investigation into what matters to learners language Massive Open Online Courses (LMOOCs). The study conducted content and sentiment analyses learner reviews LMOOCs investigate key themes/subthemes in learners' emotional tendencies as well differences detected themes/subthemes. dataset for this included 8,671 collected from 42 College English MOOCs hosted at a major MOOC platform China. analysis identified four themes—course, instructor,...

Large-scale image datasets frequently contain unavoidable noisy labels, resulting in overfitting deep neural networks and declining performance. Most existing methods for learning from labels operate as one-stage frameworks, where training data division semi-supervised (SSL) are intertwined optimization. Accordingly, their effectiveness is significantly influenced by the precision of separated clean set, prior knowledge noise, robustness SSL. In this paper, we propose a progressive sample...

Deep neural networks suffer from overfitting when training samples contain inaccurate annotations (noisy labels), leading to suboptimal performance. In addressing this challenge, current methods for learning with noisy labels employ specific criteria, such as small loss, historical prediction, etc., distinguish clean and instances. Subsequently, semi-supervised techniques are introduced boost Most of them one-stage frameworks that aim achieve optimal sample partitioning robust SSL within a...

Deep neural networks (DNNs) have achieved breakthrough progress in various fields, largely owing to the support of large-scale datasets with manually annotated labels. However, obtaining such is costly and time-consuming, making high-quality annotation a challenging task. In this work, we propose an improved noisy sample selection method, termed “sample framework”, based on mixup loss recalibration strategy (SMR). This framework enhances robustness generalization abilities models. First,...