检索结果-内蒙古大学图书馆

Parallel Accurate Minifloat MACCs for Neural Network Inference on Versal FPGAs

IEEE TRANSACTIONS ON computeR-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS 2025年第6期44卷 2181-2194页

作者： Damsgaard, Hans Jakob Hosfeld, Konstantin J. Nurmi, Jari Preuser, Thomas B. Tampere Univ SoC Hub Res Ctr Tampere 33100 Finland Tampere Univ Wireless Res Ctr Tampere 33100 Finland AMD Res & Adv Dev D-01067 Dresden Germany Stanford Univ Elect Engn Dept Stanford CA 94305 USA

Machine learning (ML) is ubiquitous in contemporary applications. Its need for efficient acceleration has driven vast research efforts into the quantization of neural networks with low-precision numerical formats. Models quantized with minifloat formats of eight or fewer bits have proven capable of outperforming models quantized into same-size integers. However, unlike integers, minifloats require accurate accumulation to prevent the introduction of rounding errors. We explore the design space of parallel accurate minifloat multiply-accumulators (MACCs) targeting the AMD VersalTM FPGA fabric. We experiment with three variations of the multiply-and-shift and adder tree components of a minifloat MACC. For comparison, we apply similar alterations to a parallel integer MACC. Our results show that custom compressor trees with external sign-inversion gates reduce the mean area of the minifloat MACCs by 17.7% and increase their clock frequency by 16.2%. In comparison, custom compressor trees with absorbed partial product generation gates reduce the mean area of integer MACCs by 28.1% and increase their clock frequency by 3.60%. Comparing the best-performing designs, we observe that minifloat MACCs consume 20% to 180% more resources than integer ones with same-size operands without accounting for a conversion back into a floating-point format, and 60% to 300% more resources when including it. Our data enable engineers to make informed decisions in their designs of deeply integrated embedded ML solutions when trading off training and fine-tuning effort versus resource cost.

关键词： Table lookup Field programmable gate arrays Quantization (signal) Accuracy Computational modeling Logic gates Hardware Design automation Costs Adders Field-programmable gate array (FPGA) floating-point arithmetic machine learning (ML) multiply-accumulate

来源：评论

学校读者我要写书评

暂无评论

Optimizing Force Fields with Experimental Data Using Ensemble Reweighting and Potential Contrasting

引用

JOURNAL OF PHYSICAL CHEMISTRY B 2024年第28期128卷 6760-6769页

作者： Ding, Xinqiang Tufts Univ Dept Chem 62 Talbot Ave Medford MA 02155 USA

Despite force field improvements over the past decades, we still encounter situations where simulation results disagree with experiments due to force field inaccuracies. Such situations provide opportunities to improve force fields. In this study, we introduce a novel framework for optimizing force fields using experimental data. The unique feature of this framework is that it aims to optimize force fields to match experiments while minimizing the perturbation made to the original force field. To achieve this, we combine ensemble reweighting techniques with the potential contrasting method. Ensemble reweighting is used to reweight an ensemble of conformations generated using an existing force field to match experimental data while minimizing the perturbation to the original ensemble. Potential contrasting is then utilized to optimize force field parameters to reproduce the reweighted ensemble. We demonstrate the framework's effectiveness by optimizing a coarse-grained force field for intrinsically disordered proteins using experimental radius of gyration data.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Network model of nociceptive processing in the superficial spinal dorsal horn reveals mechanisms of hyperalgesia, allodynia, and spinal cord stimulation

引用

JOURNAL OF NEUROPHYSIOLOGY 2023年第5期130卷 1103-1117页

作者： Gilbert, John E. Zhang, Tianhe Esteller, Rosana Grill, Warren M. Duke Univ Dept Biomed Engn Durham NC 27708 USA Boston Sci Neuromodulat Neuromodulat Researchand Adv Concepts Valencia CA USA Duke Univ Dept Elect & Comp Engn Durham NC 27708 USA Duke Univ Sch Med Dept Neurobiol Durham NC 27708 USA Duke Univ Sch Med Dept Neurosurg Durham NC 27708 USA

The spinal dorsal horn (DH) processes sensory information and plays a key role in transmitting nociception to supraspinal centers. Loss of DH inhibition during neuropathic pain unmasks a pathway from nonnociceptive A beta-afferent inputs to superficial dorsal horn (SDH) nociceptive-specific (NS) projection neurons, and this change may contribute to hyperalgesia and allodynia. We developed and validated a computational model of SDH neuronal circuitry that links nonnociceptive A beta-afferent inputs in lamina II/III to a NS projection neuron in lamina I via a network of excitatory interneurons. The excitatory pathway and the NS projection neuron were in turn gated by inhibitory interneurons with connections based on prior patch-clamp recordings. Changing synaptic weights in the computational model to replicate neuropathic pain states unmasked a low-threshold excitatory pathway to NS neurons similar to experimental recordings. Spinal cord stimulation (SCS) is an effective therapy for neuropathic pain, and accumulating experimental evidence indicates that NS neurons in the SDH also respond to SCS. Accounting for these responses may inform therapeutic improvements, and we quantified responses to SCS in the SDH network model and examined the role of different modes of inhibitory control in modulating NS neuron responses to SCS. We combined the SDH network model with a previously published model of the deep dorsal horn (DDH) and identified optimal stimulation frequencies across different neuropathic pain conditions. Finally, we found that SCS-generated inhibition did not completely suppress model NS activity during simulated pinch inputs, providing an explanation of why SCS does not eliminate acute pain.

关键词： computation model pain spinal cord

来源：评论

学校读者我要写书评

暂无评论

Meta-analysis and partial correlation coefficients: A matter of weights

引用

RESEARCH SYNTHESIS METHODS 2024年第2期15卷 303-312页

作者： Hong, Sanghyun Reed, W. Robert Univ Canterbury Dept Econ & Finance Christchurch New Zealand Univ Canterbury UCMeta Christchurch New Zealand Univ Canterbury Dept Econ & Finance & UCMeta Christchurch New Zealand

This study builds on the simulation framework of a recent paper by Stanley and Doucouliagos (Research Synthesis Methods 2023;14;515-519). S&D use simulations to make the argument that meta-analyses using partial correlation coefficients (PCCs) should employ a "suboptimal" estimator of the PCC standard error when constructing weights for fixed effect and random effects estimation. We address concerns that their simulations and subsequent recommendation may give meta-analysts a misleading impression. While the estimator they promote dominates the "correct" formula in their Monte Carlo framework, there are other estimators that perform even better. We conclude that more research is needed before best practice recommendations can be made for meta-analyses with PCCs.

关键词： bias mean squared errors meta-analysis partial correlation coefficients

来源：评论

学校读者我要写书评

暂无评论

An improved theory of thermoelasticity for ultrafast heating of materials using short and ultrashort laser pulses

引用

INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER 2023年第1期215卷

作者： Zarei, Alireza Pilla, Srikanth Clemson Univ Dept Automot Engn Greenville SC 29607 USA Clemson Univ Clemson Composites Ctr Greenville SC 29607 USA Clemson Univ Dept Mat Sci & Engn Clemson SC 29634 USA Clemson Univ Dept Mech Engn Clemson SC 29634 USA

This paper introduces an improved thermoelastic theory designed specifically to address the challenges posed by fast and ultrafast heating of materials. The proposed theory incorporates a time delay in the temperature gradient and introduces the rate of temperature gradient as an additional term in the heat equation. This term accounts for the phase lag and the precedence of the temperature gradient and heat flux vector. By leveraging the sign of the temperature gradient rate, the theory automatically determines the precedence of either the temperature gradient or the heat flux vector. To verify the effectiveness of the proposed theory, experiments reported in the literature were simulated using several theories, namely the classical, Lord-Schulman (LS), Green-Lindsay (GL), dual-phase-lag (DPL), and the presented theory. The simulations considered the effects of relaxation times and pulse duration on the results. The findings reveal that for pulse durations shorter than 50 ps, the second-order time derivative of the temperature gradient becomes a crucial term in the energy equation for accurately predicting temperature profiles. Additionally, as the pulse duration decreases, the sensitivity to relaxation times increases. Regarding the surface temperature corresponding to the material ablation threshold, it is observed that for an 80 fs laser pulse, the ablation temperature falls between the boiling and critical temperatures, significantly higher than the boiling temperature associated with the ablation by a 270 ps pulse duration. Lastly, the simulation results exhibit a high level of agreement with the experimental data, particularly in terms of displacements, further validating the accuracy of the presented simulations.

关键词： Theory of thermoelasticity Laser ultrasonics Short laser pulse Ultrafast heating Numerical simulation

来源：评论

学校读者我要写书评

暂无评论

A single-nucleus and spatial transcriptomic atlas of the COVID-19 liver reveals topological, functional, and regenerative organ disruption in patients

引用

GENOME BIOLOGY 2025年第1期26卷 1-46页

作者： Pita-Juarez, Yered Karagkouni, Dimitra Kalavros, Nikolaos Melms, Johannes C. Niezen, Sebastian Delorey, Toni M. Essene, Adam L. Brook, Olga R. Pant, Deepti Skelton-Badlani, Disha Naderi, Pourya Huang, Pinzhu Pan, Liuliu Hether, Tyler Andrews, Tallulah S. Ziegler, Carly G. K. Reeves, Jason Myloserdnyy, Andriy Chen, Rachel Nam, Andy Phelan, Stefan Liang, Yan Gregory, Mark He, Shanshan Patrick, Michael Rane, Tushar Wardhani, Aster Amin, Amit Dipak Biermann, Jana Hibshoosh, Hanina Veregge, Molly Kramer, Zachary Jacobs, Christopher Yalcin, Yusuf Phillips, Devan Slyper, Michal Subramanian, Ayshwarya Ashenberg, Orr Bloom-Ackermann, Zohar Tran, Victoria M. Gomez, James Sturm, Alexander Zhang, Shuting Fleming, Stephen J. Warren, Sarah Beechem, Joseph Hung, Deborah Babadi, Mehrtash Padera Jr, Robert F. Macparland, Sonya A. Bader, Gary D. Imad, Nasser Solomon, Isaac H. Miller, Eric Riedel, Stefan Porter, Caroline B. M. Villani, Alexandra-Chloe Tsai, Linus T. -Y. Hide, Winston Szabo, Gyongyi Hecht, Jonathan Rozenblatt-Rosen, Orit Shalek, Alex K. Izar, Benjamin Regev, Aviv Popov, Yury V. Jiang, Z. Gordon Vlachos, Ioannis S. Beth Israel Deaconess Med Ctr Dept Pathol Boston MA 02215 USA Harvard Med Sch Boston MA 02115 USA Broad Inst MIT & Harvard Cambridge MA 02142 USA Beth Israel Deaconess Med Ctr Spatial Technol Unit HMS Initiat RNA Med Pathol Boston MA 02215 USA Columbia Univ Irving Med Ctr Dept Med Div Hematol Oncol New York NY 10032 USA Columbia Ctr Translat Immunol New York NY 08807 USA Beth Israel Deaconess Med Ctr Dept Med Boston MA 02215 USA Beth Israel Deaconess Med Ctr Div Gastroenterol Hepatol & Nutr Boston MA 02215 USA Broad Inst MIT & Harvard Klarman Cell Observ Cambridge MA 02142 USA Beth Israel Deaconess Med Ctr Div Endocrinol Diabet & Metab Boston MA 02115 USA Boston Nutr & Obes Res Ctr Funct Genom & Bioinformat Core Boston MA USA Beth Israel Deaconess Med Ctr Dept Radiol Boston MA USA NanoString Technol Inc Seattle WA USA Univ Hlth Network Toronto Gen Hosp Res Inst Ajmera Transplant Ctr Toronto ON Canada Harvard Med Sch Harvard MIT Program Hlth Sci & Technol Boston MA 02115 USA MIT Boston MA USA MIT Inst Med Engn & Sci Cambridge MA 02139 USA MIT Koch Inst Integrat Canc Res Cambridge MA 02139 USA Ragon Inst MGH MIT & Harvard Cambridge MA 02139 USA Harvard Univ Harvard Grad Program Biophys Cambridge MA 02138 USA Harvard Stem Cell Inst Cambridge MA 02138 USA MIT Program Computat & Syst Biol Cambridge MA 02139 USA Harvard Med Sch Program Immunol Boston MA 02115 USA MIT Dept Chem Cambridge MA 02139 USA Columbia Univ Irving Med Ctr Dept Pathol & Cell Biol New York NY 10032 USA Genetech 1 DNA Way South San Francisco CA 94080 USA Broad Inst MIT & Harvard Infect Dis & Microbiome Program Cambridge MA 02142 USA Broad Inst MIT & Harvard Data Sci Platform Cambridge MA 02142 USA Broad Inst MIT & Harvard Precis Cardiol Lab Cambridge MA 02142 USA Harvard Med Sch Dept Genet Boston MA 02115 USA Massachusetts Gen Hosp Dept Mol Biol Boston MA 02114 USA Massachusetts Gen Hosp Ctr Computat & Integrat Biol Boston MA USA Brigham & Womens Hosp Dep

BackgroundThe molecular underpinnings of organ dysfunction in severe COVID-19 and its potential long-term sequelae are under intense investigation. To shed light on these in the context of liver function, we perform single-nucleus RNA-seq and spatial transcriptomic profiling of livers from 17 COVID-19 *** identify hepatocytes positive for SARS-CoV-2 RNA with an expression phenotype resembling infected lung epithelial cells, and a central role in a pro-fibrotic TGF beta signaling cell-cell communications network. Integrated analysis and comparisons with healthy controls reveal extensive changes in the cellular composition and expression states in COVID-19 liver, providing the underpinning of hepatocellular injury, ductular reaction, pathologic vascular expansion, and fibrogenesis characteristic of COVID-19 cholangiopathy. We also observe Kupffer cell proliferation and erythrocyte progenitors for the first time in a human liver single-cell atlas. Despite the absence of a clinical acute liver injury phenotype, endothelial cell composition is dramatically impacted in COVID-19, concomitantly with extensive alterations and profibrogenic activation of reactive cholangiocytes and mesenchymal *** atlas provides novel insights into liver physiology and pathology in COVID-19 and forms a foundational resource for its investigation and understanding.

关键词： SARS-CoV-2 COVID-19 Spatial transcriptomics Single-nucleus sequencing Liver Single-cell sequencing

来源：评论

学校读者我要写书评

暂无评论

PointNet-based modeling of systematic distance deviations for improved TLS accuracy

引用

JOURNAL OF APPLIED GEODESY 2024年第4期18卷 613-628页

作者： Hartmann, Jan Ernst, Dominik Neumann, Ingo Alkhatib, Hamza Leibniz Univ Hannover Geodet Inst Hannover Germany

Terrestrial laser scanners (TLSs) have become indispensable for acquiring highly detailed and accurate 3D representations of the physical world. However, the acquired data is subject to systematic deviations in distance measurements due to external influences, such as distance and incidence angle. This research introduces a calibration approach by applying a deep learning model based on PointNet to predict and correct these systematic distance deviations, incorporating not only the XYZ coordinates but also additional features like intensity, incidence angle, and distances within a local neighbourhood radius of 5 cm. By predicting and subsequently correcting systematic distance deviations, the quality of TLS point clouds can be improved. Hence, our model is designed to complement and build upon the foundation of prior internal TLS calibration. A data set collected under controlled environmental conditions, containing various objects of different materials, served as the basis for training and validation the PointNet based model. In addition our analysis showcase the model's capability to accurately model systematic distance deviations, outperforming existing methods like gradient boosting trees by capturing the spatial relationships and dependencies within the data more effectively. By defining test data sets, excluded from the training process, we underscore the ongoing effectiveness of our model's distance measurement calibration, showcasing its ability to improve the accuracy of the TLS point cloud.

关键词： terrestrial laser scanning systematic distance deviation deep learning PointNet calibration

来源：评论

学校读者我要写书评

暂无评论

Unexpected Two-Stage Swelling of Weak Polyelectrolyte Brushes with Divalent Counterions

引用

ACS MACRO LETTERS 2024年第9期13卷 1185-1191页

作者： Beyer, David Holm, Christian Univ Stuttgart Inst Computat Phys D-70569 Stuttgart Germany

We use particle-based, coarse-grained simulations to study the influence of divalent counterions on a weak polyelectrolyte brush. Our simulations show a profound influence of even small concentrations of divalent salt on the titration behavior of the brush, which is shown to be a combined effect of electrostatic interactions and the Donnan effect. Furthermore, we examine the partitioning of mono- and divalent counterions into the brush. We demonstrate the preferred uptake of divalent ions by the brush, which is further enhanced by electrostatic correlation effects. Finally, our simulations reveal a hitherto unobserved two-stage swelling of the brush as a function of the pH in the presence of divalent salt. This phenomenon arises as a consequence of charge regulation and ion partitioning.

关键词： Electrostatics

来源：评论

学校读者我要写书评

暂无评论

How Charge Regulation Affects pH-Responsive Ampholyte Uptake in Weak Polyelectrolyte Brushes

引用

MACROMOLECULES 2024年第1期58卷 389-402页

作者： Radhakrishnan, Keerthi Beyer, David Holm, Christian Univ Stuttgart Inst Computat Phys D-70569 Stuttgart Germany

Weak polyelectrolyte brushes are a promising platform for the selective capture and release of charged proteins from bulk solutions. Despite their potential for creating smart responsive surfaces, a detailed microscopic understanding of the uptake behavior in dependence of various parameters remains elusive. In this study, we employ coarse-grained, particle-based simulations to investigate how charge regulation under varying environmental conditions modulate the uptake and release of pH-responsive ampholytes, serving as a toy model for proteins, into weak polyelectrolyte brushes. For quenched brushes with constant ionization, the uptake of ampholytes remains strong across different isoelectric points. In contrast, for weak brushes, the ampholyte uptake becomes selectively sensitive to different isoelectric points and pK(A)-values and exhibits a nonmonotonic behavior with changing pH. Enhanced proton partitioning into the brush lowers the local pH, significantly shifting the ionization states of both the brush (pK(A)(app) > pK(A)) and ampholytes (pI(app) > pI), such that the concurrent ionization of the brush and the ampholyte results in an optimum uptake strength for pK(A)(app) < pH < pI(app). Adjusting the salt concentration broadens the uptake window and shifts the maximum uptake to higher pH values. Additionally, ampholytes with higher charge regulation capacitance near the isoelectric point demonstrate stronger adsorption, extending selective adsorption capabilities in ampholyte mixtures with similar isoelectric points.

关键词： pH

来源：评论

学校读者我要写书评

暂无评论

Interpretable Molecular Property Predictions Using Marginalized Graph Kernels

引用

JOURNAL OF CHEMICAL INFORMATION AND MODELING 2023年第15期63卷 4633-4640页

作者： Xiang, Yan Tang, Yu-Hang Lin, Guang Reker, Daniel Duke Univ Dept Biomed Engn Durham NC 27705 USA Lawrence Berkeley Natl Lab Computat Res Div Berkeley CA 94720 USA Purdue Univ Dept Math W Lafayette IN 47907 USA Purdue Univ Sch Mech Engn W Lafayette IN 47907 USA

Marginalized graph kernels have shown competitive performanceinmolecular machine learning tasks but currently lack measures of interpretability,which are important to improve trust in the models, detect biases,and inform molecular optimization campaigns. We here conceive andimplement two interpretability measures for Gaussian process regressionusing a marginalized graph kernel (GPR-MGK) to quantify (1) the contributionof specific training data to the prediction and (2) the contributionof specific nodes of the graph to the prediction. We demonstrate theapplicability of these interpretability measures for molecular propertyprediction. We compare GPR-MGK to graph neural networks on four logicand two real-world toxicology data sets and find that the atomic attributionof GPR-MGK generally outperforms the atomic attribution of graph neuralnetworks. We also perform a detailed molecular attribution analysisusing the FreeSolv data set, showing how molecules in the trainingset influence machine learning predictions and why Morgan fingerprintsperform poorly on this data set. This is the first systematic examinationof the interpretability of GPR-MGK and thereby is an important stepin the further maturation of marginalized graph kernel methods forinterpretable molecular predictions.

关键词： Algorithms Machine learning Mathematical methods Molecular modeling Molecules

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：