A5010085296- Adversarial Robustness in Machine Learning
- Topic Modeling
- Anomaly Detection Techniques and Applications
- Natural Language Processing Techniques
- Semantic Web and Ontologies
- Advanced Data Storage Technologies
- Data Management and Algorithms
- Advanced Malware Detection Techniques
- Data Quality and Management
- Image Processing Techniques and Applications
- Gut microbiota and health
- Immune Response and Inflammation
- CCD and CMOS Imaging Sensors
- Explainable Artificial Intelligence (XAI)
- VLSI and Analog Circuit Testing
- Semiconductor materials and devices
- Cervical Cancer and HPV Research
- Synthesis and properties of polymers
- Infrared Target Detection Methodologies
- Cellular Automata and Applications
- Bone Metabolism and Diseases
- IL-33, ST2, and ILC Pathways
- Eosinophilic Esophagitis
- Multimodal Machine Learning Applications
- Conducting polymers and applications
Chonnam National University
2008-2025
Samsung (South Korea)
2011-2024
Pohang University of Science and Technology
2013-2021
Incheon National University
2021
Emory University
2020
IBM (United States)
2018
Yonsei University
2015-2016
Korea Post
2013-2014
Catholic University of Daegu
2005-2013
Microsoft Research Asia (China)
2011-2013
While machine learning (ML) models are being increasingly trusted to make decisions in different and varying areas, the safety of systems using such has become an increasing concern. In particular, ML often trained on data from potentially untrustworthy sources, providing adversaries with opportunity manipulate them by inserting carefully crafted samples into training set. Recent work shown that this type attack, called a poisoning allows insert backdoors or trojans model, enabling malicious...
The expression of matrix metalloproteinases (MMPs) has been implicated in the invasion and metastasis cancer cells. Here we examined effect ascochlorin, a prenyl-phenol anti-tumor compound from fungus Ascochyta viciae, on regulation signaling pathways that control MMP-9 human renal carcinoma (Caki-1) Ascochlorin reduced invasive activity Caki-1 cells inhibited phorbol 12-myristate 13-acetate-induced increases dose-dependent manner. Reporter gene, electrophoretic mobility shift, kinase...
Machine learning architectures are readily available, but obtaining the high quality labeled data for training is costly. Pre-trained models available as cloud services can be used to generate this costly data, and would allow an attacker replicate trained models, effectively stealing them. Limiting information provided by based omitting class probabilities has been proposed a means of protection significantly impacts utility models. In work, we illustrate how still provide useful...
Large language models (LLMs) are vulnerable to universal jailbreaks-prompting strategies that systematically bypass model safeguards and enable users carry out harmful processes require many interactions, like manufacturing illegal substances at scale. To defend against these attacks, we introduce Constitutional Classifiers: trained on synthetic data, generated by prompting LLMs with natural rules (i.e., a constitution) specifying permitted restricted content. In over 3,000 estimated hours...
Cancer cells, characterized by local invasion and distant metastasis, are very much dependant on the extracellular matrix. The expression of matrix metalloproteinases (MMPs) has been implicated in metastasis cancer cells. In this study, we reported effects disulfiram, a clinically used anti-alcoholism drug, tumor suppression, as well its activity MMP-2 MMP-9 human osteosarcoma cells (U2OS). Disulfiram for alcohol aversion therapy. However, recent reports have shown that disulfiram may...
A monolithic 64 Gb MLC NAND flash based on 21 <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">$~$</tex></formula> nm process technology has been developed. The device consists of 4-plane arrays and provides page size up to 32 KB. It also features a newly developed asynchronous DDR interface that can support the maximum bandwidth 400 MB/s. To improve performance reliability, on-chip randomizer, soft...
Knowledge bases, which consist of concepts, entities, attributes and relations, are increasingly important in a wide range applications. We argue that knowledge about (of concepts or entities) plays critical role inferencing. In this paper, we propose methods to derive for millions quantify the typicality with regard their corresponding concepts. employ multiple data sources such as web documents, search logs, existing scores by aggregating different distributions derived from using methods....
3D-NAND Flash memory is evaluated as a key device that handling the explosive data growth, showing steady bit growth and areal density increases >30% every year. In particular, 4b/cell (QLC) technology appearing to be one of most promising solutions meet market demand with excellent cost effectiveness, while QLC being gradually adopted by various storage applications [1]. However, there are several challenges need addressing so rapidly market: (1) reprogram on-chip buffered program (OBP)...
The extended latissimus dorsi flap has been widely used for breast reconstruction. However, seroma at the donor site is a common complication and makes it difficult reconstructive surgeons to choose as primary option. We analyzed association between reconstructions with flaps. A series of 120 consecutive cases were included in this study. average body mass index (BMI) was 22.1 kg/m, mean ratio weight extirpated equivalent 101.6%. age patients 40.3 years. Donor-site reported 69.2% (83 cases)...
The market growth of mobile applications such as smart phones and tablet computers has fueled the explosive demand NAND Flash memories having high density fast throughput. To meet a demand, we present 64Gb multilevel cell (MLC) memory 533Mb/s DDR interface in sub-20nm technology. Large floating-gate (FG) coupling interference program disturbance are major challenges to impede scaling technology node [1]. In this paper, correction-before-coupling (CBC) reprogram P3-pattern pre-pulse scheme,...
Arpita Roy, Youngja Park, Taesung Lee, Shimei Pan. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.
Interleukin-10 plays important, yet contrasting, roles in host protection against bacterial infections and the septic response. To determine role of IL-10 defense Acinetobacter baumannii infection, wild-type (WT) IL-10-deficient mice were infected intranasally with bacteria. exhibited increased mortality, severe pathology, excess production proinflammatory cytokines chemokines lungs, burdens bronchoalveolar lavage (BAL) fluids lung homogenates after A. compared to WT mice. Intranasal...
Deep neural networks (DNNs) have been widely applied in various applications, including autonomous driving and surveillance systems. However, DNNs are found to be vulnerable adversarial examples, which carefully crafted inputs aiming mislead a learner make incorrect predictions. While several defense detection approaches proposed for static image classification, many security-critical tasks use videos as their input require efficient processing. In this paper, we propose an effective method...
Helicobacter pylori is a gram-negative, microaerophilic, and spiral-shaped bacterium causes gastrointestinal diseases in human. IL-1β representative cytokine produced innate immune cells considered to be key factor the development of malignancies. However, mechanism production by neutrophils during H. infection still unknown. We designed this study identify host bacterial factors involved regulation pylori-induced neutrophils. found that abolished NLRP3-, ASC-, caspase-1/11-deficient...
Atopic dermatitis (AD) is one of the most prevalent, chronic and persistent inflammatory skin diseases closely associated with intestinal microbiota. To evaluate effect D-galactose intake on AD, we orally administered to BALB/c mice whose ears were treated 2,4-dinitrochlorobenzene (DNCB). alleviated DNCB-induced AD-like phenotypes such as redness, scaling/dryness excoriation. Ear thickness was also decreased by administration. Histopathological analysis revealed epidermal thickening,...
Inflammatory bowel disease (IBD) is an intestinal chronic inflammatory disease, and its incidence steadily increasing. IBD closely related to the microbiota, probiotics are known be a potential therapeutic agent for IBD. In our study, we evaluated protective effect of Lactobacillus sakei CVL-001, isolated from Baechu kimchi, on dextran sulfated sodium (DSS)-induced colitis in mice. The oral administration L. CVL-001 according experimental schedule alleviated weight loss activity mice with...
Large ontologies and taxonomies are automatically harvested from web-scale data. These tend to be huge, noisy, contains little context. As a result, cleansing enriching those large-scale becomes great challenge. A natural way enrich taxonomy is map the existing datasets that contain rich information. In this paper, we study problem of matching two web scale taxonomies. Besides problem, address challenge may not enough context (such as attribute values). entity resolution techniques based...
Important cloud services rely on spatial-keyword queries, containing a spatial predicate and arbitrary boolean keyword queries. In particular, we study the processing of such queries in main memory to support short response times. contrast, current state-of-the-art indexes relational engines are designed for different assumptions. Rather than building new index, employ cost-based optimizer process these using index index. We address several technical challenges achieve this goal. introduce...