To collect large scale annotated data, it is inevitable to introduce label noise, i.e., incorrect class labels. be robust against many successful methods rely on the noisy classifiers (i.e., models trained training data) determine whether a trustworthy. However, remains unknown why this heuristic works well in practice. In paper, we provide first theoretical explanation for these methods. We prove that prediction of classifier can indeed good indicator data clean. Based result, propose novel...

In recent years, graph-based deep learning algorithms have attracted widespread attention in the field of consumer electronics. Still, most current graph neural networks are based on supervised or semi-supervised learning, which often relies true labels given samples as auxiliary information. To solve this problem, we propose a Deep Self-Supervised Attention Convolution Autoencoder Graph Clustering (DSAGC) model and use it for social clustering. We divide proposed into two parts: pretext...

Renal carcinoma is a frequently seen cancer globally, with laparoscopic partial nephrectomy (LPN) being the primary form of treatment. Accurately identifying renal structures such as kidneys, tumors, veins, and arteries on CT scans crucial for optimal surgical preparation However, automatic segmentation these remains challenging due to kidney's complex anatomy variability imaging data. This study presents RenalSegNet, novel deep-learning framework automatically segmenting structure in...

In this paper, we propose a new clustering-based binary-class classification framework that integrates the clustering technique into approach to handle imbalanced data sets. A classifier is designed classify set of instances two classes; while partitions groups according their similarity each other. After applying algorithm, within same group usually have higher similarity, and differences among between different should be larger. our proposed framework, all negative are first clustered...

Recent years, transfer learning has attracted much attention in the community of machine learning. In this paper, we mainly focus on tasks parameter under framework extreme (ELM). Unlike existing approaches, which incorporate source model information into target by regularizing di erence between and domain parameters, an intuitively appealing projective-model is proposed to bridge parameters. Specifically, formulate ELM networks means projection, train optimizing projection matrix classifier...

Localized scanning is a simple technique used by attackers to search for vulnerable hosts. trades off between the local and global of hosts has been Code Red II Ninida worms. As such strategy so yet effective in attacking Internet, it important that defenders understand spreading ability behaviors localized-scanning In this work, we first characterize relationships vulnerable-host distributions spread worms through mathematical modeling analysis, compare random with localized scanning. We...

Traffic classification is a fundamental component in advanced network management and security. Recent research has achieved certain success the application of machine learning techniques into flow statistical feature based approach. However, most methods classify traffic on assumption that all flows are generated by known applications. Considering pervasive unknown applications real world environment, this does not hold. In paper, we cast as specific problem with insufficient negative...

This paper presents a new semi-supervised method to effectively improve traffic classification performance when few supervised training data are available. Existing semi methods label large proportion of testing flows as unknown due limited information, which severely affects the performance. To address this problem, we propose incorporate flow correlation into both and stages. At stage, make use extend set by automatically labeling unlabeled according their pre-labeled flows. Consequently,...

Reinforcement learning (RL) has achieved promising results in solving numerous challenging sequential decision problems. To address the issue of sparse extrinsic rewards, researchers have proposed intrinsic enabling agent to acquire skills that may prove valuable pursuit future rewards. One representative approach for generating rewards involves constructing a predictive model assess novelty states. However, due stochastic nature complex environments, can be noisy. Directly employing noisy...

In this paper, we present our work balancing ontological and operational factors in building collaborations within multiagent neighborhoods. This innovation takes into account the desired level of performance, service priorities, relaying tasks to determine whether an agent should entertain learning, which are more expensive but rewarding long run, or carry out less short term. The domain application is multiagent, distributed information retrieval, where agents, safe-guarding data...

Deep learning classifiers for characterization of whole slide tissue morphology require large volumes annotated data to learn variations across different and cancer types. As is well known, manual generation digital pathology training time consuming expensive. In this paper, we propose a semi-automated method annotating group similar instances at once, instead collecting only per-instance annotations. This allows much larger set, that reflects visual variability multiple types thus single...

Objective The precise segmentation of kidneys from a 2D ultrasound (US) image is crucial for diagnosing and monitoring kidney diseases. However, achieving detailed difficult due to US images’ low signal-to-noise ratio low-contrast object boundaries. Methods This paper presents an approach called deep supervised attention with multi-loss functions (MLAU-Net) segmentation. MLAU-Net model combines the benefits mechanisms supervision improve accuracy. mechanism allows selectively focus on...

Recently, considerable effort has been devoted to deep domain adaptation in computer vision and machine learning communities. However, most of existing work only concentrates on shared feature representation by minimizing the distribution discrepancy across different domains. Due fact that all alignment approaches can reduce, but not remove shift. Target samples distributed near edge clusters, or far from their corresponding class centers are easily be misclassified hyperplane learned source...

Recently, webly supervised learning (WSL) has been studied to leverage numerous and accessible data from the Internet. Most existing methods focus on noise-robust models web images while neglecting performance drop caused by differences between domain real-world domain. However, only tackling gap above can we fully exploit practical value of datasets. To this end, propose a Few-shot guided Prototypical (FoPro) representation method, which needs few labeled examples reality significantly...

A critical problem for Internet traffic classification is how to obtain a high-performance statistical feature based classifier using small set of training data. The solutions this are essential deal with the encrypted applications and new emerging applications. In paper, we propose Naive Bayes (NB) scheme tackle problem, which utilizes two recent research findings, discretization flow correlation. bag-of-flow (BoF) model firstly introduced describe correlated flows it leads BoF-based...

The past decade has seen a lot of research on statistics-based network protocol identification using machine learning techniques. Prior studies have shown promising results in terms high accuracy and fast classification speed. However, most works embodied an implicit assumption that all protocols are known advance presented the training data, which is unrealistic since real-world networks constantly witness emerging traffic patterns as well unknown wild. In this paper, we revisit problem by...

The integration of large language models (LLMs) with social robots has emerged as a promising avenue for enhancing human-robot interactions at time when news reports generated by artificial intelligence (AI) are gaining in credibility. This is expected to intensify and become more productive resource journalism, media, communication, education. In this paper novel system proposed that integrates AI's generative pretrained transformer (GPT) model the Pepper robot, aim improving robot's...

Probabilistic inferences distill knowledge from graphs to aid human make important decisions. Due the inherent uncertainty in model and complexity of knowledge, it is desirable help end-users understand inference outcomes. Different deep or high dimensional parametric models, lack interpretability graphical models due cyclic long-range dependencies byzantine procedures. Prior works did not tackle cycles interpretable. We formulate explanation probabilistic as a constrained cross-entropy...