检索结果-内蒙古大学图书馆

Efficient and scalable neural residual waveform coding with collaborative quantization

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Zhen, Kai Lee, Mi Suk Sung, Jongmo Beack, Seungkwon Kim, Minje Indiana University Luddy School of Informatics Computing and Engineering BloomingtonIN United States Indiana University Cognitive Science Program BloomingtonIN United States Electronics and Telecommunications Research Institute Daejeon Korea Republic of

Scalability and efficiency are desired in neural speech codecs, which supports a wide range of bitrates for applications on various devices. We propose a collaborative quantization (CQ) scheme to jointly learn the codebook of LPC coefficients and the corresponding residuals. CQ does not simply shoehorn LPC to a neural network, but bridges the computational capacity of advanced neural network models and traditional, yet efficient and domain-specific digital signal processing methods in an integrated manner. We demonstrate that CQ achieves much higher quality than its predecessor at 9 kbps with even lower model complexity. We also show that CQ can scale up to 24 kbps where it outperforms AMR-WB and Opus. As a neural waveform codec, CQ models are with less than 1 million parameters, significantly less than many other generative models. Copyright © 2020, The Authors. All rights reserved.

关键词： Deep neural networks

Precision Rehabilitation for Patients Post-Stroke based on Electronic Health Records and Machine Learning

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Gao, Fengyi Zhang, Xingyu Sivarajkumar, Sonish Denny, Parker E. Aldhahwani, Bayan M. Visweswaran, Shyam Shi, Ryan Hogan, William Bove, Allyn Wang, Yanshan Department of Health Information Management University of Pittsburgh PittsburghPA United States Department of Communication Science and Disorders University of Pittsburgh PittsburghPA United States Intelligent Systems Program School of Computing and Information University of Pittsburgh PittsburghPA United States Department of Physical Therapy University of Pittsburgh PittsburghPA United States Department of Medical Rehabilitation Sciences Umm Al-Qura University Makkah Saudi Arabia Department of Biomedical Informatics University of Pittsburgh PittsburghPA United States Department of Computer Science University of Pittsburgh Medical Center PittsburghPA United States Data Science Institute Medical College of Wisconsin MilwaukeeWI United States Hillman Cancer Center University of Pittsburgh Medical Center PittsburghPA United States

Objective In this study, we utilized statistical analysis and machine learning methods to examine whether rehabilitation exercises can improve patients post-stroke functional abilities, as well as forecast the improvement in functional abilities. Our dataset is patients’ rehabilitation exercises and demographic information recorded in the unstructured electronic health records (EHRs) data and free-text rehabilitation procedure notes. Through this study, our ultimate goal is to pinpoint the specific rehabilitation exercises that can effectively aid post-stroke patients in improving their functional outcomes in basic mobility (BM) and applied cognitive (AC) domains. Data sources We collected data for 265 stroke patients from the University of Pittsburgh Medical Center, accessed through the Rehabilitation Datamart With informatics iNfrastructure for Research (ReDWINE). Methods We employed a pre-existing natural language processing (NLP) algorithm to extract data on rehabilitation exercises and developed a rule-based NLP algorithm to extract Activity Measure for Post-Acute Care (AM-PAC) scores, covering basic mobility (BM) and applied cognitive (AC) domains, from procedure notes. AM-PAC scores were collected at the initial rehabilitation visit and followed up at one and two months—key recovery periods. Changes in AM-PAC scores were classified based on the minimal clinically important difference (MCID), and significance was assessed using Friedman and Wilcoxon tests. To identify impactful exercises, we used Chi-square tests, Fisher's exact tests, and logistic regression for odds ratios. Additionally, we developed five machine learning models—logistic regression (LR), Adaboost (ADB), support vector machine (SVM), gradient boosting (GB), and random forest (RF)—to predict outcomes in functional ability. Results Statistical analyses revealed significant associations between functional improvements and specific exercises. In the AC domain, the BALANCE exercise showed substant

关键词： Statistical tests

Knowledge-Guided Multiview Deep Curriculum Learning for Elbow Fracture Classification

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Luo, Jun Kitamura, Gene Arefan, Dooman Doganay, Emine Panigrahy, Ashok Wu, Shandong Intelligent Systems Program School of Computing and Information University of Pittsburgh PittsburghPA United States Department of Radiology School of Medicine University of Pittsburgh PittsburghPA United States University of Pittsburgh Medical Center Children’s Hospital of Pittsburgh PittsburghPA United States Department of Biomedical Informatics Department of Bioengineering University of Pittsburgh PA United States

Elbow fracture diagnosis often requires patients to take both frontal and lateral views of elbow X-ray radiographs. In this paper, we propose a multiview deep learning method for an elbow fracture subtype classification task. Our strategy leverages transfer learning by first training two single-view models, one for frontal view and the other for lateral view, and then transferring the weights to the corresponding layers in the proposed multiview network architecture. Meanwhile, quantitative medical knowledge was integrated into the training process through a curriculum learning framework, which enables the model to first learn from "easier" samples and then transition to "harder" samples to reach better performance. In addition, our multiview network can work both in a dual-view setting and with a single view as input. We evaluate our method through extensive experiments on a classification task of elbow fracture with a dataset of 1,964 images. Results show that our method outperforms two related methods on bone fracture study in multiple settings, and our technique is able to boost the performance of the compared methods. The code is available at https://***/ljaiverson/multiview-curriculum. Copyright © 2021, The Authors. All rights reserved.

关键词： Curricula

Synthetic biology in Indonesia: Potential and projection in a country with mega biodiversity

学校读者我要写书评

暂无评论

Biotechnology Notes 2023年第1期4卷 41-48页

作者： Sanka, Immanuel Kusuma, Ali Budhi Martha, Faustina Hendrawan, Andre Pramanda, Ihsan Tria Wicaksono, Adhityo Jati, Afif Pranaya Mazaya, Maulida Dwijayanti, Ari Izzati, Nurul Maulana, Muhammad Farhan Widyaningrum, Aulia Reski Synthetic Biology Indonesia (Synbio.id) Jl. Raya Lintas Sumbawa-Bima Block AA No. 1 Boak Village Subdistrict Unter Iwes Sumbawa 84316 Indonesia Department of Chemistry and Biotechnology Tallinn University of Technology Akadeemia tee 15 Tallinn 12618 Estonia Indonesian Centre for Extremophile Bioresources and Biotechnology (ICEBB) Faculty of Life Sciences and Technology Sumbawa University of Technology Jl. Raya Olat Maras Sumbawa 84371 Indonesia Science Communication Steering Committee iGEM Foundation 45 Prospect St Cambridge 02139 MA United States Department of Bio Technology Indonesia International Institute for Life Sciences (i3L) Jl. Pulomas Barat Kav. 88 Pulomas Jakarta 13210 Indonesia Division of Biotechnology Genbinesia Foundation Jalan Swadaya Barat no. 4 Gresik 61171 Indonesia Infection Program Monash Biomedicine Discovery Institute and Department of Microbiology Monash University Melbourne VIC Australia Research Center for Computing Research Organization for Electronics and Informatics National Research and Innovation Agency (BRIN) Cibinong Science Center Jl. Raya Jakarta-Bogor KM 46 West Java Cibinong 16911 Indonesia

Synthetic biology has gained many interest around the globe in the last two decades, not only due to its rapid development but also the potential to provide addressable solutions using standardized design of biological systems. Considering its huge population, biodiversity, and natural resources, Indonesia could play an important role in shaping the future of synthetic biology towards a sustainable bio-circular economy. Here, we provide an overview of synthetic biology development in Indonesia, especially on exploring the potential of our biodiversity. We also discuss some potentials of synthetic biology in solving national issues. Furthermore, we also provide the projection and future landscape of synthetic biology development in Indonesia. In addition, we briefly explain the potential challenges that may arise during the development. © 2023 The Authors

关键词： Indonesia Mega biodiversity Synbio development Synthetic biology

Unveiling hepatitis B biomarkers through genomic network analysis

学校读者我要写书评

暂无评论

Egyptian Journal of Medical Human Genetics 2025年第1期26卷 1-11页

作者： Adikusuma, Wirawan Firdayani, Firdayani Irham, Lalu Muhammad Santri, Ichtiarini Nurullita Phiri, Yohane Vincent Abero Pitaloka, Dian Ayu Eka Chong, Rockie Sarasmita, Made Ary Satria, Rahmat Dani Research Center for Computing Research Organization for Electronics and Informatics National Research and Innovation Agency (BRIN) Cibinong Indonesia Department of Pharmacy Universitas Muhammadiyah Mataram Mataram Indonesia Research Center for Vaccine and Drugs Research Organization for Health National Research and Innovation Agency (BRIN) South Tangerang Indonesia Faculty of Pharmacy Universitas Ahmad Dahlan Yogyakarta Indonesia Faculty of Public Health Unversitas Ahmad Dahlan Yogyakarta Indonesia Department of Epidemiology and Environmental Health (EEH) University at Buffalo Buffalo USA Department of Pharmacology and Clinical Pharmacy Faculty of Pharmacy Universitas Padjadjaran Sumedang Indonesia Center of Excellence in Higher Education for Pharmaceutical Care Innovation Universitas Padjadjaran Sumedang Indonesia Department of Chemistry and Biochemistry University of California Los Angeles USA Pharmacy Study Program Faculty of Mathematics and Sciences Udayana University Bali Indonesia Department of Clinical Pathology and Laboratory Medicine Faculty of Medicine Public Health and Nursing Universitas Gadjah Mada Yogyakarta Indonesia Integrated Laboratory Installation Dr. Sardjito General Hospital Yogyakarta Indonesia

The infection caused by the Hepatitis B virus (HBV) poses a significant challenge to global health. However, identifying and prioritizing functional biomarkers associated with the disease will allow more focused efforts for pre-clinical investigation. This study aimed to employ advanced genomic networks and bioinformatics tools to prioritize biomarkers for HBV. Comprehensive curation from six well-known databases yields a set of 495 HBV-associated genes based on gene ontology enrichment, pathway enrichment and protein–protein interaction analyses. Further functional gene network analyses identified 10 hub genes (IL-6, TNF, STAT3, CD4, IL-10, STAT1, IL-1B, AKT1, IL-2, and TLR4), warranting further investigation. Among these genes, IL-6, TNF, and STAT3 emerged as promising biomarkers for HBV through CytoHubba analysis. Overall, this study offers a genomic network-centric approach to identify and prioritize potential HBV biomarkers to provide valuable insights into its pathogenesis.

关键词：

Deep curriculum learning in task space for multi-class based mammography diagnosis

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Luo, Jun Arefan, Dooman Zuley, Margarita Sumkin, Jules Wu, Shandong Intelligent Systems Program School of Computing and Information University of Pittsburgh PittsburghPA United States Department of Radiology University of Pittsburgh PittsburghPA United States Magee-Womens Hospital University of Pittsburgh Medical Center PittsburghPA United States Department of Biomedical Informatics University of Pittsburgh PittsburghPA United States Department of Bioengineering University of Pittsburgh PittsburghPA United States

Mammography is used as a standard screening procedure for the potential patients of breast cancer. Over the past decade, it has been shown that deep learning techniques have succeeded in reaching near-human performance in a number of tasks, and its application in mammography is one of the topics that medical researchers most concentrate on. In this work, we propose an end-to-end Curriculum Learning (CL) strategy in task space for classifying the three categories of Full-Field Digital Mammography (FFDM), namely Malignant, Negative, and False recall. Specifically, our method treats this three-class classification as a "harder" task in terms of CL, and create an "easier" sub-task of classifying False recall against the combined group of Negative and Malignant. We introduce a loss scheduler to dynamically weight the contribution of the losses from the two tasks throughout the entire training process. We conduct experiments on an FFDM datasets of 1,709 images using 5-fold cross validation. The results show that our curriculum learning strategy can boost the performance for classifying the three categories of FFDM compared to the baseline strategies for model training. Copyright © 2021, The Authors. All rights reserved.

关键词： Curricula

VLSI Design of Tree-Based Inference for Low-Power Learning Applications

学校读者我要写书评

暂无评论

VLSI Design of Tree-Based Inference for Low-Power Learning A...

IEEE International Symposium on Circuits and Systems (ISCAS)

作者： Brunno A. Abreu Mateus Grellert Sergio Bampi Graduate Program in Microelectronics (PGMICRO) INF-UFRGS Federal University of Rio Grande do Sul Porto Alegre Brazil Department of Informatics and Statistics − INE Federal University of Santa Catarina (UFSC) Florianópolis Brazil Graduate Program on Electronic Engineering and Computing Catholic University of Pelotas (UCPel) Pelotas Brazil

ISBN: (数字)9781728133201

ISBN: (纸本)9781728133218

The use of Machine Learning techniques in battery-powered devices has increased in recent years. Therefore, evaluating power-accuracy trade-offs of inference models has become important. This work explores decision trees architectures, analyzing the effects of model complexity and approximation in power and accuracy. By quantizing the inputs to limited widths, we increased the accuracy of the models in up to 8.7% compared to precise cases. In terms of power, the variation in model complexity was more significant than in input width, as we obtained reductions of up to 97% by decreasing the tree depth, compared to 88% when decreasing the width.

关键词： Complexity theory Hardware Integrated circuit modeling Training Analytical models Decision trees Quantization (signal)

A dual-mode deep transfer learning (D2TL) system for breast cancer detection using contrast enhanced digital mammograms

学校读者我要写书评

暂无评论

IISE Transactions on Healthcare Systems Engineering 2019年第4期9卷 357-370页

作者： Wang, Kun Patel, Bhavika K. Wang, Lujia Wu, Teresa Zheng, Bin Li, Jing Industrial Engineering Program at School of Computing Informatics and Decision Systems Engineering Arizona State University Tempe United States Department of Radiology Mayo Clinic in Arizona Scottsdale United States School of Electrical and Computer Engineering College of Engineering University of Oklahoma Norman United States

Full-field digital mammography (FFDM) and magnetic resonance imaging (MRI) are gold-standard techniques for breast cancer detection. The newly contrast-enhanced digital mammography (CEDM) integrates the complementary strengths of FFDM and MRI, and is being incorporated into the practice of leading institutions. The current clinical practice using CEDM is sub-optimal because it is primarily based on clinicians' trained eyes. Automated diagnostic systems under the conventional machine learning paradigm suffer from drawbacks, such as the requirement for precise segmentation, extraction of shallow features that do not suffice for diagnostic images, and adoption of a sequential design without a global objective. We propose a deep learning (DL)-empowered diagnostic system using CEDM, the core of which is a novel dual-mode deep transfer learning (D2TL) model. The proposed system is innovative in several aspects, including (1) a dual-mode deep architecture design;(2) use of transfer learning to facilitate robust model estimation under small sample-size;(3) development of visualization techniques to help interpret the model results and facilitate inter- and intra-tumor malignancy quantification;and (4) minimization of human bias. We apply D2TL to classify benign vs. malignant tumors using the CEDM data collected from the Mayo Clinic in Arizona. D2TL outperforms competing models and approaches. © 2019, © 2019 “IISE”.

关键词： breast cancer Deep learning imaging-based diagnosis transfer learning

Low-Power and Memory-Aware Approximate Hardware Architecture for Fractional Motion Estimation Interpolation on HEVC

学校读者我要写书评

暂无评论

Low-Power and Memory-Aware Approximate Hardware Architecture...

IEEE International Symposium on Circuits and Systems (ISCAS)

作者： Wagner Penny Guilherme Correa Luciano Agostini Daniel Palomino Marcelo Porto Gabriel Nazar Bruno Zatt Video Technology Research Group − Graduate Program in Computing − Federal University of Pelotas − Pelotas Brazil Sul-rio-grandense Federal Institute − Pelotas Brazil Institute of Informatics − Federal University of Rio Grande do Sul − Porto Alegre Brazil

ISBN: (数字)9781728133201

ISBN: (纸本)9781728133218

Nowadays, current video coding standards like the High Efficiency Video Coding (HEVC) implement several complex coding tools, like the Fractional Motion Estimation (FME). An alternative to improve performance and save power is allying the hardware acceleration with approximate computing solutions, focusing on such complex tools. In this work, we present a low-power and memory-aware hardware architecture for the HEVC FME interpolator, proposing the development of two novel hardware designs for the interpolation filters, called Approximate Unified FME Filters (AUFF). These solutions exploit the usage of approximate computing at both algorithmic and data levels, leading to a reduction in dissipated power and memory bandwidth. The proposed design is capable of real-time interpolation of UHD (Ultra High Definition) 4K and 8K videos when synthesized using a 40 nm standard-cell library, with a power dissipation ranging from 22.04 to 62.06 mW.

关键词： Finite impulse response filters Hardware Interpolation Encoding Computer architecture Approximate computing Motion estimation