Semi-supervised learning (SSL) techniques such as Label Propagation (LP) and Label Spreading (LS) leverage a small number of labeled samples together with large unlabeled datasets. However, these graph-based methods often perform poorly on tabular data, where explicit graph structures are absent and feature relationships are heterogeneous. Conventional adaptations using k-nearest neighbor graphs or RBF kernels suffer from instability, information loss, and sensitivity to label noise. This study proposes Kernel Discriminant Analysis with Uncertainty Modeling (GF-KDA), a label refinement framework that models pairwise similarities directly in the feature space without relying on graph construction. The method incorporates feature-wise uncertainty to reduce the influence of unreliable attributes and employs a posterior-based confidence mechanism to control noisy label diffusion. Experiments on three public tabular benchmarks—Adult Income, Heart Disease, and Credit Risk—demonstrate that GF-KDA consistently achieves higher refinement accuracy and stability across varying label noise ratios (0–50%) and labeled data fractions (10–30%). The results highlight GF-KDA as a noise-resilient and computationally efficient approach for semi-supervised learning in tabular domains.

Tabular data, a cornerstone of real-world applications in domains such as healthcare, finance, and infrastructure, pose unique challenges for deep learning due to their heterogeneous feature types, lack of spatial structure, and frequent presence of noisy or missing values. Although recent advances in self-supervised learning (SSL) have enabled progress in tabular representation learning, existing methods often overlook a critical aspect of real-world data—feature-level uncertainty. In this paper, we propose UA-Tab (Uncertainty-Aware self-supervised representation learning for Tabular data), a novel self-supervised framework that explicitly models and leverages uncertainty in tabular data. UA-Tab introduces two key innovations: (1) a unified attention-uncertainty mechanism that identifies and emphasizes informative yet reliable features while down-weighting noisy or ambiguous inputs, and (2) a latent-space perturbation strategy for view generation, enabling contrastive learning without the need for unreliable negative sampling. Through comprehensive experiments on the Adult Income dataset under a wide range of synthetic noise scenarios—including value corruption, masking, mode collapse, and out-of-distribution anomalies—UA-Tab demonstrates superior robustness and generalization compared to existing SSL baselines such as VIME, SCARF, STaB, SubTab, and TabDeCo. Notably, UA-Tab consistently achieves the highest accuracy and macro F1-score across conditions while maintaining interpretability through its uncertainty outputs. These results highlight UA-Tab’s potential to serve as a reliable and transparent foundation for self-supervised tabular learning, especially in environments where data quality varies and explainability is critical.

Pose-Guided Person Image Synthesis (PGPIS) aims to generate human images in specified poses while preserving the identity and appearance of a source image. This technology facilitates diverse applications, including virtual try-on, digital avatars, animation, and sign language generation. Despite the high-quality results of recent diffusion-based PGPIS, these models typically depend on implicit feature aggregation within the denoising process. As a result, fine-grained texture preservation is limited, and even for the same identity, it is difficult to ensure consistent generation under variations in pose and source appearance. To address these limitations, we propose Fusion Embedding for PGPIS using a Diffusion Model (FPDM), the first framework that explicitly aligns fused source–pose embeddings with target image embeddings via contrastive learning and subsequently employs the learned fusion embedding as a conditioning signal for generation. FPDM integrates an Image–Pose Fusion (IPF) module into our proposed Source-Enhanced Pose Fusion approach to learn a fusion embedding aligned with the target image. We then employ a conditional diffusion model guided by source appearance, target pose, and the learned fusion embedding. Experiments on the DeepFashion benchmark and the RWTH-PHOENIX-Weather 2014T dataset demonstrate competitive performance compared to existing methods in both quantitative and qualitative evaluations, with ablation studies confirming that explicit fusion embedding alignment substantially improves texture fidelity and consistency across pose and source appearance variations. The implementation is publicly available at https://github.com/dhlee-work/FPDM .

The stability and user trust in cloud services depends on prompt detection and response to diverse anomalies. This study focuses on an Ensemble-based anomaly detection methodology that integrates log data with computing resource metrics, aiming to overcome the limitations of traditional single-data models. To process the unstructured nature of log data, we use the Drain Parser to transform it into a structured format, and Doc2Vec embeds it. The study adheres to a reconstruction-based approach for anomaly detection, specifically building upon the Anomaly Transformer model. The proposed model leverages the concept of an Anomaly Transformer based on the Attention mechanism. It integrates preprocessed metric data with log data for effective anomaly detection. Experiments were conducted using metric and log data collected from real-world cloud environments. The model’s performance was evaluated based on accuracy, recall, precision, f1 score, and AUROC. The results demonstrate that our proposed Ensemble-based model outperforms traditional models such as LSTM, VAR, and deeplog.

While everyday document search is done by keyword-based queries to search engines, we have situations that need deep search of documents such as scrutinies of patents, legal documents, and so on. In such cases, using document queries, instead of keyword-based queries, can be more helpful because it exploits more information from the query document. This paper studies a scheme of document search based on document queries. In particular, it uses centrality vectors, instead of tf-idf vectors, to represent query documents, combined with the Word2vec method to capture the semantic similarity in contained words. This scheme improves the performance of document search and provides a way to find documents not only lexically, but semantically close to a query document.

Research and development projects have close relationship with laws. In some cases, new technologies resulted from R&D projects can't be used because some statutes restrict them. The reason of this problem is that researchers don't know exactly which laws can affect their R&D projects. To solve the issue, we suggest a model for law retrieval system that can be used by researchers of R&D projects to find related statutes. Input of this model is a query document that describes the main contents of a project. By using ontology, legal terms are extracted from the document and statutes defining them are retrieved as a set of related laws. After this searching process, statutes are provided to researchers with their ranks, which are assigned using relevance scores we developed. By using this model, we can make a system for researchers to search a list of statutes that may affect R&D projects, and finally, they can adjust their project's direction by checking the list, preventing their works from being useless.

Supervised learning require sentiment labeled corpus for training. But it is hard to apply automatic sentiment classification system to new domain because labeled dataset construction costs a lot of time. Meanwhile, researches using Doc2vec based document representation beat out other sentiment classification researches. However, these document representation methods only represent documents' context or sentiment. In this paper, we proposed supervised learning scheme for unlabeled corpus and also proposed document representation method which can simultaneously represent documents' context and sentiment.

This paper has analyzed the Centrality measures on individual keywords hierarchy in telecommunication business in Bangladesh and South Korea, which are significantly correlated with BDTelecom company and SKTelecom Company have the most dominant position in both networks. The division and membership of the clusters in both networks also showed some similarities and comparisons in the global flow of telecom business.

We address the problem of specifying Web searches and retrieving, filtering and rating Web pages so as to improve the relevance and quality of hits, based on the user's search intent and preferences. We present a methodology and architecture for an agent-based system, called WebSifter II, that captures the semantics of a user's decision-oriented search intent, transforms the semantic query into target queries for existing search engines, and then ranks the resulting page hits according to a user-specified weighted-rating scheme. Users create personalized search taxonomies via our weighted semantic-taxonomy tree. Consulting a Web taxonomy agent such as Wordnet helps refine the terms in the tree. The concepts represented in the tree are then transformed into a collection of queries processed by existing search engines. Each returned page is rated according to user-specified preferences such as semantic relevance, syntactic relevance, categorical match, page popularity and authority/hub rating.