检索结果-内蒙古大学图书馆

IEEE International Conference on Big data

作者： Nicholas Josselyn Biao Yin Ziming Zhang Elke Rundensteiner Department of Data Science Worcester Polytechnic Institute Worcester MA Department of Electrical and Computer Engineering Worcester Polytechnic Institute Worcester MA Department of Computer Science Worcester Polytechnic Institute Worcester MA

ISBN: (纸本)9781665480468

Deep learning often relies on the availability of a large amount of high-quality labeled data, which can be very limited in novel domains. To address such data scarcity, domain adaptation is one promising approach that allows for deep networks to leverage large amounts of available data from a source domain to enhance the model’s efficacy on the target domain of interest. However, while there is a plethora of alternate models for domain adaptation proposed over many years in the literature, there is a dearth of studies that objectively compare the relative effectiveness of these models in a rigorous, empirical study. To fill this gap, we provide a thorough, unbiased, empirical study of five state-of-the-art (SOTA) deep domain adaptation models proposed over the past 6 years whose codes are publicly available. Models are evaluated on the complex and diverse domain adaptation tasks featured in the DomainNet benchmark dataset as well as the popular Office-31 dataset. Our results suggest that (1) all 5 models perform similarly, on average, and do not even significantly beat the oldest model, and (2) counter to their intended purpose, the transfer loss functions in the literature do not contribute significantly to learning transferable representations. Our observations suggest that domain adaptation research needs to more thoroughly compare newly proposed models against existing works, along with assessing their loss functions’ utility thoroughly. Our code and data splits are made public for reproducibility of results by the community.

关键词： Deep learning Adaptation models Codes Big data Benchmark testing Reproducibility of results data models

来源：评论

学校读者我要写书评

暂无评论

Depression Diagnosis: Bioinformatics Uncovers Cellular Abnormalities in Female-Specific Genes with Ai-Driven Blood Biomarker-Based Prediction 24

Depression Diagnosis: Bioinformatics Uncovers Cellular Abnor...

引用

24th IEEE International Conference on Bioinformatics and Bioengineering, BIBE 2024

作者： Firoz, Neda Yadav, Nivedita Beresteneva, Olga Grigorievna Vladimirovich, Sergey Aksyonov Institute of Applied Mathematics and Computer Science Tomsk State Univeristy Tomsk Russia Indian Biological Data Center Regional Centre for Biotechnology Faridabad India Tomsk Polytechnic Univeristy Tomsk Russia

ISBN: (纸本)9798331518622

With an estimated 264 million people affected by MDD globally, there is an urgent need for novel approaches to diagnosis and treatment. This study explored the complex gene expression profiles linked to Major Depressive Disorder (MDD) using state-of-the-art Next-Generation Sequencing (NGS), namely RNA-Seq. There is a need for accurate, non-invasive diagnostics, particularly through gender-specific biomarker research in female RNA-Seq data. To address this disparity, RNA-Seq data was analyzed using Machine Learning (ML) techniques to categorize female-specific MDD, identifying 10 critical biomarkers and utilizing Artificial Intelligence (AI) and bioinformatics to extract transcriptome data from publicly available datasets, while employing Generative Adversarial Networks (GANs) for data augmentation due to the limited availability of large RNA-Seq datasets. Gene set enrichment analysis (GSEA) and other bioinformatics analyses further demonstrated that genes differentially expressed in MDD patients were enriched in pathways related to the cell cycle, neutrophil degranulation, resolution of sister chromatid cohesion, and the formation of mitotic spindles by EML4 and NUDC. Ten important biomarkers associated with MDD were identified by an integrative bioinformatics investigation: IGHV3-64D, IGHV5-10-1, IFI27, LILRB5, PBK, SIGLEC1, IFI44L, CDCA5, and SLC4A1. Machine learning models based on these genes achieved 95% accuracy, highlighting potential of RNA-Seq and ML to enhance MDD diagnosis and reveal its molecular mechanisms. © 2024 IEEE.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

RevPRAG: Revealing Poisoning Attacks in Retrieval-Augmented Generation through LLM Activation Analysis

arXiv

引用

arXiv 2024年

作者： Tan, Xue Luan, Hao Luo, Mingyu Sun, Xiaoyan Chen, Ping Dai, Jun School of Computer Science Fudan University Shanghai China Department of Computer Science Worcester Polytechnic Institute MA United States

Retrieval-Augmented Generation (RAG) enriches the input to LLMs by retrieving information from the relevant knowledge database, enabling them to produce responses that are more accurate and contextually appropriate. It is worth noting that the knowledge database, being sourced from publicly available channels such as Wikipedia, inevitably introduces a new attack surface. RAG poisoning attack involves injecting malicious texts into the knowledge database, ultimately leading to the generation of the attacker’s target response (also called poisoned response). However, there are currently limited methods available for detecting such poisoning attacks. We aim to bridge the gap in this work by introducing RevPRAG, a flexible and automated detection pipeline that leverages the activations of LLMs for poisoned response detection. Our investigation uncovers distinct patterns in LLMs’ activations when generating poisoned responses versus correct responses. Our results on multiple benchmarks and RAG architectures show our approach can achieve a 98% true positive rate, while maintaining a false positive rate close to 1%. © 2024, CC BY-SA.

关键词： database systems

来源：评论

学校读者我要写书评

暂无评论

Machine Learning Prediction of Just Dance Exergame Enjoyment from Mobile Sensor data

Machine Learning Prediction of Just Dance Exergame Enjoyment...

引用

IEEE Symposium on Computational Intelligence and Games, CIG

作者： Joshua Audibert Elijah Gonzalez Ryan Orlando Nicholas Wong Emmanuel Agu Mark Claypool Computer Science Department Worcester Polytechnic Institute (WPI) Worcester MA USA

Many young adults do not exercise enough, choosing instead to spend time on electronic media (e.g., smartphone, Internet). Exergames, which gamify physical activity, have been shown to be effective at increasing physical activity in an enjoyable way. For exergames to remain effective, sustained user engagement is key. However, sustaining long-term engagement in games (including exergames) is a challenging research problem – 95% of all new game players stop playing within 3 months, and 85% of new players stop after just one day. We posit that if detected early, waning player exergame enjoyment can be countered by recommending new, more enjoyable games before the player quits playing. In this paper, we investigate machine learning to predict user enjoyment of the Just Dance exergame by analyzing data gathered from the player’s smartphone. Specifically, "ground truth" scores for the players’ enjoyment obtained from the Immersive Experience Questionnaire (IEQ) E-scores are inferred from user behaviors such as increased excitement and gameplay frequency. These, in turn, are predicted by data gathered by the phone’s sensors – accelerometer, gyroscope and game features. Analysis of data from a user study shows the Naive Bayes classification algorithm achieves the best results, achieving 75% accuracy for binary classification (enjoying vs. not enjoying the exergame) of enjoyment E-scores. The most predictive features were the energy in the 0.5 to 3 GHz range, windowed energy in the 0.5 to 3 Hz range and radio spectral peak using a Discrete Cosine Transform (DCT). Our results are preliminary but encouraging and we plan to improve on our results by collecting more data and utilizing state-of-the-art neural networks approaches.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Optimized Teammate Selection with a Focus on Member Compatibility to Domain Roles

SSRN

引用

SSRN 2023年

作者： Galbiati, Federico Gran, Ranier X. Jacques, Brendan D. Mulhern, Sullivan J. Ngan, Chun-Kit Department of Computer Science Worcester Polytechnic Institute United States

The goal of this project is to develop a Machine Learning tool for predicting the performance of a team of individuals based solely on the strengths and weaknesses of its members in past experiences prior to the team’s formation. There exist models in the field for predicting an existing team’s performance based on that team’s past performance but having a model that helps optimize team formation can assist and improve existing processes further. We tested our theory by developing a tool to assist with the drafting process in FIRST Robotics competitions, which involves the top eight competitors in each tournament selecting their teammates from the other competitors based on their performance in a series of qualification matches. We developed a model to evaluate individual and team strengths and weaknesses, optimize team formation based on the model, and perform game winner prediction with 80% validation accuracy. © 2023, The Authors. All rights reserved.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

MediatorDNN: Contention Mitigation for Co-Located DNN Inference Jobs

MediatorDNN: Contention Mitigation for Co-Located DNN Infere...

引用

IEEE International Conference on Cloud Computing, CLOUD

作者： Seyed Morteza Nabavinejad Sherief Reda Tian Guo Computer Science Department Worcester Polytechnic Institute Worcester MA School of Engineering Brown University Providence RI

ISBN: (数字)9798350368536

ISBN: (纸本)9798350368543

With the increase in computing power of cutting-edge hardware platforms, it is a common practice to run multiple jobs on a single machine for improved resource utilization and throughput. However, this leads to inevitable resource contention among co-located jobs, impacting their performance. The resource contention can worsen due to fluctuations in resource utilization of jobs caused by variations in their input workload. To tackle the co-location contention for DNN inference jobs, we propose MediatorDNN, which considers contention and resource utilization variation when co-locating DNN inference jobs. It profiles each DNN, monitors microarchitectural metrics such as memory bandwidth and cache access pattern, and high-level resource utilization like CPU utilization. Based on profiling results and leveraging Modern Portfolio Theory (MPT), MediatorDNN decides on the co-location of jobs. Experimental results with various DNNs on two hardware platforms show that MediatorDNN improves throughput by up to 108% (21 % on average) compared to an approach only considering contention and ignoring resource utilization variation.

关键词： Measurement Cloud computing Microarchitecture Prevention and mitigation Artificial neural networks Throughput Hardware

来源：评论

学校读者我要写书评

暂无评论

FairCIM: Fair Interference Mitigation by DNN Switching for Latency-Sensitive Inference Jobs

FairCIM: Fair Interference Mitigation by DNN Switching for L...

引用

Autonomic Computing and Self-Organizing Systems (ACSOS),International Conference on

作者： Seyed Morteza Nabavinejad Sherief Reda Tian Guo Computer Science Department Worcester Polytechnic Institute Worcester MA School of Engineering Brown University Providence RI

ISBN: (数字)9798350363876

ISBN: (纸本)9798350363883

It is a common practice to co-locate several jobs, such as DNN inference ones, on a single machine to improve resource utilization, energy efficiency, and overall performance. However, individual job performance might degrade when these jobs compete for shared resources. This degradation can in particular affect latency-sensitive jobs bound by Service Level Agreements (SLA), and result in SLA violations. To tackle this issue and mitigate the impact of interference on latency, we design a system called FairCIM. For certain applications such as image classification, various DNNs with varying accuracy and computational complexity are proposed. Upon detection of SLA violations, FairCIM dynamically switches between those DNNs to maintain the latency in the presence of co-location interference. To ensure fairness, FairCIM selects the jobs for DNN switching based on their resource utilization. We conduct experiments using diverse DNNs, datasets, and hardware platforms to evaluate the FairCIM effectiveness. Compared to another approach that uses a resource exchange mechanism, FairCIM’s flexibility achieved by DNN switching substantially reduces SLA violations and improves the effective accuracy by up to 244%.

关键词： Accuracy Prevention and mitigation Artificial neural networks Switches Interference Hardware Resource management Faces Service level agreements Image classification

来源：评论

学校读者我要写书评

暂无评论

Toward Intelligent Adaptive Airborne GPR, Implementation and data Acquisition

Toward Intelligent Adaptive Airborne GPR, Implementation and...

引用

IEEE International Conference on Wireless for Space and Extreme Environments (WiSEE)

作者： Saeed Haghniaz Jahromi Vincent Filardi Seyed Reza Zekavat Zhonghai Wang Joshua Thurber Dylan Hoffman Charlotte Larson Douglas Petkie Department of Physics Worcester Polytechnic Institute Worcester MA USA Department of Data Science Worcester Polytechnic Institute Worcester MA USA Chief Technology Office Alex and Alice Inc. MD USA

ISBN: (数字)9798350351118

ISBN: (纸本)9798350351125

This paper is the first attempt toward the realization of intelligent soil subsurface moisture characterization via a pulse-modulated Stepped Frequency Continuous Wave (SFCW) Ground Penetrating Radar (GPR). The GPR design is standalone, autonomous, and drone-rated (compact and light-weight), facilitating the development of an airborne GPR for megafarm moisture assessment. The design also enables the implementation of diverse waveforms for the transmitted signal. Thus, the waveform can adaptively be adjusted to optimally extract the soil subsurface features from the received signal. Based on a measurement campaign conducted in the summer of 2024, the paper assesses the performance of moisture estimation using multiple Machine Learning models applied to the received signal power features across all frequency components of the SFCW waveform. The results confirm the potential of the designed radar for intelligent soil moisture estimation.

关键词： Wireless communication Moisture measurement Soil measurements Soil moisture Measurement uncertainty Moisture Estimation Feature extraction Frequency estimation Probes

来源：评论

学校读者我要写书评

暂无评论

Local Treewidth of Random and Noisy Graphs with Applications to Stopping Contagion in Networks 25

Local Treewidth of Random and Noisy Graphs with Applications...

引用

25th International Conference on Approximation Algorithms for Combinatorial Optimization Problems and the 26th International Conference on Randomization and Computation, APPROX/RANDOM 2022

作者： Mehta, Hermish Reichman, Daniel Citadel Securities ChicagoIL United States Department of Computer Science Worcester Polytechnic Institute MA United States

ISBN: (纸本)9783959772495

We study the notion of local treewidth in sparse random graphs: the maximum treewidth over all k-vertex subgraphs of an n-vertex graph. When k is not too large, we give nearly tight bounds for this local treewidth parameter;we also derive nearly tight bounds for the local treewidth of noisy trees, trees where every non-edge is added independently with small probability. We apply our upper bounds on the local treewidth to obtain fixed parameter tractable algorithms (on random graphs and noisy trees) for edge-removal problems centered around containing a contagious process evolving over a network. In these problems, our main parameter of study is k, the number of initially "infected" vertices in the network. For the random graph models we consider and a certain range of parameters the running time of our algorithms on n-vertex graphs is 2o(k) poly(n), improving upon the 2Ω(k) poly(n) performance of the best-known algorithms designed for worst-case instances of these edge deletion problems. © Hermish Mehta and Daniel Reichman.

关键词： Graphic methods

来源：评论

学校读者我要写书评

暂无评论

Robot-Assisted Wide-Area Photoacoustic System

Robot-Assisted Wide-Area Photoacoustic System

引用

IEEE Symposium (IUS) Ultrasonics

作者： Shang Gao Xihan Ma Haichong K. Zhang Department of Robotics Engineering Worcester Polytechnic Institute Worcester USA Department of Biomedical Engineering Worcester Polytechnic Institute Worcester USA Department of Computer Science Worcester Polytechnic Institute Worcester USA

Liver ischemia and reperfusion injury (IRI) is a vascular complication that can lead to hepatic functional impairment. IRI is relevant in liver transplantation and during liver surgery performed with intermittent vascular inflow occlusion. Detecting intraoperative liver oxygenation impairment and ischemia could help prevent IRI, which is challenging partially due to the multiple hepatic vascular inflows. We propose a robot-assisted wide-area photoacoustic (PA) imaging system to monitor the liver oxygenation level intraoperatively. PA imaging has been widely used for oxygenation detection in preclinical research. We used a robotic arm to actuate the 2D PA probe and scan across the liver surface. An accurate volumetric tomography can be reconstructed via registering 2D images to the recorded robot trajectory. An ex vivo porcine liver sample was scanned using the proposed system. The result suggested that the robot-assisted wide-area PA tomography has the potential to be incorporated into clinical practice to evaluate liver SO2 intraoperatively.

关键词：

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：