Filters








4,869 Hits in 1.3 sec

Generative Single Image Reflection Separation [article]

Donghoon Lee, Ming-Hsuan Yang, Songhwai Oh
2018 arXiv   pre-print
Single image reflection separation is an ill-posed problem since two scenes, a transmitted scene and a reflected scene, need to be inferred from a single observation. To make the problem tractable, in this work we assume that categories of two scenes are known. It allows us to address the problem by generating both scenes that belong to the categories while their contents are constrained to match with the observed image. A novel network architecture is proposed to render realistic images of
more » ... scenes based on adversarial learning. The network can be trained in a weakly supervised manner, i.e., it learns to separate an observed image without corresponding ground truth images of transmission and reflection scenes which are difficult to collect in practice. Experimental results on real and synthetic datasets demonstrate that the proposed algorithm performs favorably against existing methods.
arXiv:1801.04102v1 fatcat:oaqh4j6uizdjvl4t7o3546yuvy

ODSQA: Open-domain Spoken Question Answering Dataset [article]

Chia-Hsuan Lee and Shang-Ming Wang and Huan-Cheng Chang and Hung-Yi Lee
2018 arXiv   pre-print
Reading comprehension by machine has been widely studied, but machine comprehension of spoken content is still a less investigated problem. In this paper, we release Open-Domain Spoken Question Answering Dataset (ODSQA) with more than three thousand questions. To the best of our knowledge, this is the largest real SQA dataset. On this dataset, we found that ASR errors have catastrophic impact on SQA. To mitigate the effect of ASR errors, subword units are involved, which brings consistent
more » ... ements over all the models. We further found that data augmentation on text-based QA training examples can improve SQA.
arXiv:1808.02280v1 fatcat:bu4i3cbha5edto346fr2mews7a

A Framework for Proactive Resource Provisioning in IaaS Clouds

Yi-Hsuan Lee, Kuo-Chan Huang, Cheng-Hsien Wu, Yen-Hsuan Kuo, Kuan-Chou Lai
2017 Applied Sciences  
Cloud computing is an emerging technology for rapidly provisioning and releasing resources on-demand from a shared resource pool. When big data is analyzed/mined on the cloud platform, the efficiency of resource provisioning would affect the system performance. This work proposes a framework for proactive resource provisioning in IaaS (Infrastructure as a Service) clouds to improve system performance. The proposed framework consists of the virtual cluster computing system, the profiling system,
more » ... the resource management system, and the monitoring system. In this framework, the over-commit mechanism is applied to improve resource utilization. Furthermore, a proactive task scheduling approach is also present to prevent the postponement of tasks in critical stages, especially when the amount of aggregated resources requested by virtual machines exceeds that of available resources on the over-committed physical machines. Experimental results show that the over-commit approach indeed improves the resource utilization. However, when the degree of applying the over-commit approach increases, the burden of this proposed approach also conceivably increases. Therefore, the proposed framework further applies the proactive task scheduling approach to execute the time-critical tasks earlier to shorten the processing time. A small-scale cloud system including 3 servers is built for experiments. Preliminary experimental results show the performance improvement of our proposed framework in IaaS clouds.
doi:10.3390/app7080777 fatcat:rrethxmbyjabpnbhrucx6vrmbi

Homomorphically Encrypted Computation using Stochastic Encodings [article]

Hsuan Hsiao, Vincent Lee, Brandon Reagen, Armin Alaghi
2022 arXiv   pre-print
Homomorphic encryption (HE) is a privacy-preserving technique that enables computation directly over ciphertext. Unfortunately, a key challenge for HE is that implementations can be impractically slow and have limits on computation that can be efficiently implemented. For instance, in Boolean constructions of HE like TFHE, arithmetic operations need to be decomposed into constituent elementary logic gates to implement so performance depends on logical circuit depth. For even heavily quantized
more » ... xed-point arithmetic operations, these HE circuit implementations can be slow. This paper explores the merit of using stochastic computing (SC) encodings to reduce the logical depth required for HE computation to enable more efficient implementations. Contrary to computation in the plaintext space where many efficient hardware implementations are available, HE provides support for only a limited number of primitive operators and their performance may not directly correlate to their plaintext performance. Our results show that by layering SC encodings on top of TFHE, we observe similar challenges and limitations that SC faces in the plaintext space. Additional breakthroughs would require more support from the HE libraries to make SC with HE a viable solution.
arXiv:2203.02547v1 fatcat:57h6cyxqhrgkdmgl6tn7mpk3uy

Proposed Diagnostic Criteria for Smartphone Addiction

Yu-Hsuan Lin, Chih-Lin Chiang, Po-Hsien Lin, Li-Ren Chang, Chih-Hung Ko, Yang-Han Lee, Sheng-Hsuan Lin, Aviv M. Weinstein
2016 PLoS ONE  
Global smartphone penetration has led to unprecedented addictive behaviors. The aims of this study are to develop diagnostic criteria of smartphone addiction and to examine the discriminative ability and the validity of the diagnostic criteria.
doi:10.1371/journal.pone.0163010 pmid:27846211 pmcid:PMC5112893 fatcat:qhc5k4cgmrherpfvbybv33rmwy

Efficient Generation of Chemically Induced Mesenchymal Stem Cells from Human Dermal Fibroblasts

Pei-Lun Lai, Hsuan Lin, Shang-Fu Chen, Shang-Chih Yang, Kuo-Hsuan Hung, Ching-Fang Chang, Hsiang-Yi Chang, Frank Leigh Lu, Yi-Hsuan Lee, Yu-Chuan Liu, Hsiao-Chun Huang, Jean Lu
2017 Scientific Reports  
Human mesenchymal stromal/stem cells (MSCs) are multipotent and currently undergoing hundreds of clinical trials for disease treatments. To date, no studies have generated induced MSCs from skin fibroblasts with chemicals or growth factors. Here, we established the first chemical method to convert primary human dermal fibroblasts into multipotent, induced MSC-like cells (iMSCs). The conversion method uses a defined cocktail of small molecules and growth factors, and it can achieve efficient
more » ... ersion with an average rate of 38% in 6 days. The iMSCs have much higher clonogenicity than fibroblasts, and they can be maintained and expanded in regular MSC medium for at least 8 passages and further differentiated into osteoblasts, adipocytes, and chondrocytes. Moreover, the iMSCs can suppress LPS-mediated acute lung injury as effectively as bone marrow-derived mesenchymal stem cells. This finding may greatly benefit stem cell biology, cell therapy, and regenerative medicine. Mesenchymal stromal/stem cells (MSCs) were first isolated from bone marrow and able to differentiate into multiple lineages, including bone, fat, cartilage, and fibroblasts 1,2 . In addition to their multipotency, MSCs are known for their immunoregulatory functions 3 and the ability to secrete multiple cytokines to promote tissue healing 4 . Currently, hundreds of clinical trials are performed to test the efficacy of MSCs in cell therapy (www.clinicaltrials. gov). According to the Guidelines of the Mesenchymal and Tissue Stem Cell Committee of the International Society for Cellular Therapy (ISCT), MSCs should fulfill three criteria, including (1) being plastic-adherent when maintained in standard culture conditions, (2) expressing CD105, CD73 and CD90 surface markers, and not expressing CD45, CD34, CD14/CD11b, CD79a/CD19 and HLA-DR surface molecules, and (3) able to differentiate into osteoblasts, adipocytes, and chondrocytes in vitro 5 . Besides, a common property of MSCs is their capacity to suppress the immune responses stimulated by endotoxin-like lipopolysaccharides (LPS) 6,7 . It is believed that MSCs exert their immunoregulatory effects by their paracrine ability to reduce inflammation and inhibit immune responses 6, 8 . Unlike many other functional cell types, due to the lack of identified tissue-specific master regulators/transcription factors in MSCs, it is hard to pinpoint a combination of key transcription factors to generate induced MSC-like cells (iMSCs) from fibroblasts. Previous studies have indicated that SSEA-4 and PODXL are two promising markers to isolate MSCs with stronger multipotency or clonogenicity 9-11 . Importantly, multipotency and clonogenicity are critical properties that clearly distinct MSCs from fibroblasts 12 . SSEA-4 high bone marrow-derived mesenchymal stem cells (BMMSCs) exhibit stronger multipotency that can differentiate into osteoblasts, chondrocytes, and adipocytes 9,10 . PODXL high BMMSCs have higher differentiation ability and proliferative ability than PODXL low BMMSCs 11 .
doi:10.1038/srep44534 pmid:28303927 pmcid:PMC5356011 fatcat:xwbgzoogpvcx5bljn7t2z3cmja

Macromolecular ion accelerator mass spectrometer

Yun-Fei Hsu, Jung-Lee Lin, Ming-Lee Chu, Yi-Sheng Wang, Chung-Hsuan Chen
2013 The Analyst  
In order to interpret the advantages to add second linear ion trap, we added four figures in the follows to explain our idea. Those four figures were captured form oscilloscope directly. The time scale of oscilloscope is shown as 100µs from Fig. 1 to 3. Time scale in figure 4 is 200µ. Then signal channel impedance of oscilloscope was set 50 ohm. Low impedance represents ion distributions. When an ion packet was ejected out from an ion trap, the time spreads were between 100 to 200µs. For longer
more » ... timing, it means the ion space distribution will longer than space between two electrodes of accelerator. The ion packet could not gain kinetic energy completely. This is why we added second linear ion trap to narrow down the ion packet. Fig. 1 The yellow color signal of cytochrome C was obtained by unstable ejected step frequency scan from 60kHz to 45kHz. The Ion trap was floated to 3000V. Electronic Supplementary Material (ESI) for Analyst This journal is
doi:10.1039/c3an01534c pmid:24171190 fatcat:rpsrnn3sp5fcfhce2wzn76i554

Flavonoids Identification and Pancreatic Beta-Cell Protective Effect of Lotus Seedpod

Ming-Shih Lee, Charng-Cherng Chyau, Chi-Ping Wang, Ting-Hsuan Wang, Jing-Hsien Chen, Hui-Hsuan Lin
2020 Antioxidants  
Oxidative stress is highly associated with the development of diabetes mellitus (DM), especially pancreatic beta-cell injury. Flavonoids derived from plants have caused important attention in the prevention or treatment of DM. Lotus seedpod belongs to a traditional Chinese herbal medicine and has been indicated to possess antioxidant, anti-age, anti-glycative, and hepatoprotective activities. The purpose of this study was to demonstrate the pancreatic beta-cell protective effects of lotus
more » ... d aqueous extracts (LSE) against oxidative injury. According to HPLC/ESI-MS-MS method, LSE was confirmed to have flavonoids derivatives, especially quercetin-3-glucuronide (Q3G). In vitro, LSE dose-dependently improved the survival and function of rat pancreatic beta-cells (RIN-m5F) from hydrogen peroxide (H2O2)-mediated loss of cell viability, impairment of insulin secretion, and promotion of oxidative stress. LSE showed potential in decreasing the H2O2-induced occurrence of apoptosis. In addition, H2O2-triggered acidic vesicular organelle formation and microtubule-associated protein light chain 3 (LC3)-II upregulation, markers of autophagy, were increased by LSE. Molecular data explored that antiapoptotic and autophagic effects of LSE, comparable to that of Q3G, might receptively be mediated via phospho-Bcl-2-associated death promoter (p-Bad)/B-cell lymphoma 2 (Bcl-2) and class III phosphatidylinositol-3 kinase (PI3K)/LC3-II signal pathway. In vivo, LSE improved the DM symptoms and pancreatic cell injury better than metformin, a drug that is routinely prescribed to treat DM. These data implied that LSE induces the autophagic signaling, leading to protect beta-cells from oxidative stress-related apoptosis and injury.
doi:10.3390/antiox9080658 pmid:32722185 fatcat:kvbjqy3nuvfmtmxhku27lel5bq

KaggleDBQA: Realistic Evaluation of Text-to-SQL Parsers [article]

Chia-Hsuan Lee, Oleksandr Polozov, Matthew Richardson
2021 arXiv   pre-print
The goal of database question answering is to enable natural language querying of real-life relational databases in diverse application domains. Recently, large-scale datasets such as Spider and WikiSQL facilitated novel modeling techniques for text-to-SQL parsing, improving zero-shot generalization to unseen databases. In this work, we examine the challenges that still prevent these techniques from practical deployment. First, we present KaggleDBQA, a new cross-domain evaluation dataset of
more » ... Web databases, with domain-specific data types, original formatting, and unrestricted questions. Second, we re-examine the choice of evaluation tasks for text-to-SQL parsers as applied in real-life settings. Finally, we augment our in-domain evaluation task with database documentation, a naturally occurring source of implicit domain knowledge. We show that KaggleDBQA presents a challenge to state-of-the-art zero-shot parsers but a more realistic evaluation setting and creative use of associated database documentation boosts their accuracy by over 13.2%, doubling their performance.
arXiv:2106.11455v1 fatcat:qx6777rqsbckzocoxna474u7nm

Unsupervised Representation Learning by Sorting Sequences [article]

Hsin-Ying Lee, Jia-Bin Huang, Maneesh Singh, Ming-Hsuan Yang
2017 arXiv   pre-print
We present an unsupervised representation learning approach using videos without semantic labels. We leverage the temporal coherence as a supervisory signal by formulating representation learning as a sequence sorting task. We take temporally shuffled frames (i.e., in non-chronological order) as inputs and train a convolutional neural network to sort the shuffled sequences. Similar to comparison-based sorting algorithms, we propose to extract features from all frame pairs and aggregate them to
more » ... redict the correct order. As sorting shuffled image sequence requires an understanding of the statistical temporal structure of images, training with such a proxy task allows us to learn rich and generalizable visual representation. We validate the effectiveness of the learned representation using our method as pre-training on high-level recognition problems. The experimental results show that our method compares favorably against state-of-the-art methods on action recognition, image classification and object detection tasks.
arXiv:1708.01246v1 fatcat:vw3mc73uobbv5ddq2locnntsje

Condensedly: comprehending article contents through condensed texts [article]

Chao-Hsuan Ke, Tsung-Lu Michael Lee, Jung-Hsien Chiang
2016 arXiv   pre-print
Summary: Abstracts in biomedical articles can provide a quick overview of the articles but detailed information cannot be obtained without reading full-text contents. Full-text articles certainly generate more information and contents; however, accessing full-text documents is usually time consuming. Condensedly is a web-based application, which provides readers an easy and efficient way to access full-text paragraphs using sentences in abstracts as fishing bait to retrieve the big fish reside
more » ... n full-text. Condensedly is based on the paragraph ranking algorithm, which evaluates and ranks full-text paragraphs based on their association scores with sentences in abstracts. Availability: http://140.116.247.185/~research/Condensedly
arXiv:1612.09062v1 fatcat:acy4uunxi5fe5egcaxw3eb3tye

Interleukin-3 Polymorphism is Associated with Miscarriage of Fresh in Vitro Fertilization Cycles

Cheng-Hsuan Wu, Tsung-Hsien Lee, Shun-Fa Yang, Hui-Mei Tsao, Yu-Jun Chang, Chia-Hsuan Chou, Maw-Sheng Lee
2019 International Journal of Environmental Research and Public Health  
Materials and Methods Study Design and Subjects Women undergoing IVF at the Lee Women's Hospital in Taichung, Taiwan (n = 1015) were prospectively enrolled in the study from January 2014 to December  ... 
doi:10.3390/ijerph16060995 pmid:30893922 pmcid:PMC6466610 fatcat:pbchzcaytnea7pjz3wbt4idtza

Lactobacillus salivarius reverse antibiotic-induced lung defense impairment in a ventilator model

Tzyy-Bin Tsay, Ming-Chieh Yang, Wan-Hsuan Chang, Pei-Hsuan Chen, Lee-Wei Chen
2018 Journal of Translational Medicine  
Widespread use of antibiotics in the intensive care unit is a potential cause of the emergence of hospital-acquired pneumonia. This study determined whether Lactobacillus salivarius feeding could reverse antibioticinduced lung defense impairment in a ventilator model. Methods: C57BL/6 wild-type (WT) mice received mechanical ventilation for 3 h after intramuscular antibiotic treatment for 6 days. Treatment with dead Lactobacillus salivarius and fructo-oligosaccharides (FOS) feeding were used to
more » ... timulate antibacterial protein expression in the intestine. Reactive oxygen species (ROS) in the intestinal mucosa was detected using 2ʹ7ʹ-dichlorofluorescein diacetate. The peroxynitrite production of alveolar macrophages (AMs) was measured using dihydrorhodamine 123 oxidation assay. N-acetylcysteine (NAC), an ROS scavenger, was orally administered to mice receiving antibiotics with FOS feeding. Results : Antibiotic treatment decreased Pseudomonas aeruginosa (PA) phagocytic activity and activity of AMs and protein expression of regenerating islet-derived protein 3β (Reg3β) as well as Toll-like receptor 4 (TLR4) in the intestinal mucosa in the ventilator model. Antibiotic treatment also decreased ROS production in the intestinal mucosa, peroxynitrite production of AMs, and RELMβ expression as well as NF-κB DNA binding activity of the intestinal mucosa in WT mice but not in MyD88 −/− mice. Treatment with dead L. salivarius or FOS feeding increased ROS production, bacterial killing activity, and protein expression of Reg3β as well as TLR4 in the intestinal mucosa and reversed the inhibitory effects of antibiotics on PA phagocytic activity of AMs. Conclusion: Taken together with the finding that ablation of FOS-induced intestinal ROS using NAC decreased peroxynitrite production as well as PA phagocytic activity of AMs and protein expression of CRP-ductin, IL-17, Reg3β, and RELMβ in the intestinal mucosa, we conclude that commensal microflora plays a key role in stimulating lung immunity. Intestinal ROS plays a role as a predictive indicator and modulator of pulmonary defense mechanisms. Antibiotic treatment reduces lung defense against PA infection through the decrease in intestinal Reg3β and TLR4 expression. Treatment with dead L. salivarius or FOS feeding reverses the antibiotic-induced lung defense impairment through the intestinal ROS/MyD88 pathways.
doi:10.1186/s12967-018-1597-1 pmid:30103798 pmcid:PMC6090719 fatcat:ipi7b64shrfjtd53xby5uo3zge

Genetic Variants of LRRK2 in Taiwanese Parkinson's Disease

Yih-Ru Wu, Kuo-Hsuan Chang, Wen-Teng Chang, Ya-Chin Hsiao, Hsuan-Chu Hsu, Pei-Ru Jiang, Yi-Chun Chen, Chih-Ying Chao, Yi-Chung Chang, Bo-Hsun Lee, Fen-Ju Hu, Wan-Ling Chen (+3 others)
2013 PLoS ONE  
Genetic variants of leucine-rich repeat kinase 2 (LRRK2) were reported to alter the risk for Parkinson's disease (PD). However, the genetic spectrum of LRRK2 variants has not been clearly disclosed yet in Taiwanese population. Herein, we sequenced LRRK2 coding region in 70 Taiwanese early onset PD patients (age at onset ≤ 50), and found six amino acid-changing single nucleotide polymorphisms (SNPs, N551K, R1398H, R1628P, S1647T, G2385R and M2397T), one reported (R1441H) and 2 novel missense
more » ... 7H and S885N) mutations. We examined the frequency of identified LRRK2 variants by genotyping 573 Taiwanese patients with PD and 503 age-matched control subjects. The results showed that PD patients demonstrated a higher frequency of G2385R A allele (4.6%) than control subjects (2.1%; odds ratio = 2.27, 95% confidence interval: 1.38-3.88, P = 0.0017). Fewer PD patients (27.7%) carried the 1647T-2397T haplotype as compared with the control subjects (33.0%; odds ratio = 0.80, 95% confidence interval: 0.65-0.97, P = 0.0215). However, the frequency of 1647T-2385R-2397T haplotype (4.3%) in PD patients was still higher than in control subjects (1.9%, odds ratio: 2.15, 95% confidence interval: 1.27-3.78, P = 0.0058). While no additional subject was found to carry R767H and R1441H, one more patient was observed to carry the S885N variant. Our results indicate a robust risk association regarding G2385R and a new possible protective haplotype (1647T-2397T). Gene-environmental interaction and a larger cohort study are warranted to validate our findings. Additionally, two new missense mutations (R767H and S885N) regarding LRRK2 in PD patients were identified. Functional studies are needed to elucidate the effects of these LRRK2 variants on protein function.
doi:10.1371/journal.pone.0082001 pmid:24339985 pmcid:PMC3855417 fatcat:tpdikdpbc5eozakzilv2c4weou

Dancing to Music [article]

Hsin-Ying Lee, Xiaodong Yang, Ming-Yu Liu, Ting-Chun Wang, Yu-Ding Lu, Ming-Hsuan Yang, Jan Kautz
2019 arXiv   pre-print
Dancing to music is an instinctive move by humans. Learning to model the music-to-dance generation process is, however, a challenging problem. It requires significant efforts to measure the correlation between music and dance as one needs to simultaneously consider multiple aspects, such as style and beat of both music and dance. Additionally, dance is inherently multimodal and various following movements of a pose at any moment are equally likely. In this paper, we propose a
more » ... is learning framework to generate dance from music. In the analysis phase, we decompose a dance into a series of basic dance units, through which the model learns how to move. In the synthesis phase, the model learns how to compose a dance by organizing multiple basic dancing movements seamlessly according to the input music. Experimental qualitative and quantitative results demonstrate that the proposed method can synthesize realistic, diverse,style-consistent, and beat-matching dances from music.
arXiv:1911.02001v1 fatcat:k4f4ty3iajhsjd4rm43lsasicu
« Previous Showing results 1 — 15 out of 4,869 results