检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Lei, Chunyu Chen, C.L. Philip Guo, Jifeng Zhang, Tong The Guangdong Provincial Key Laboratory of Computational Intelligence and Cyberspace Information The School of Computer Science and Engineering South China University of Technology Guangzhou510006 China The Brain and Affective Cognitive Research Center Pazhou Lab Guangzhou510335 China

Deep learning generally suffers from enormous computational resources and time-consuming training processes. Broad Learning System (BLS) and its convolutional variants have been proposed to mitigate these issues and have achieved superb performance in image classification. However, the existing convolutional-based broad learning system (C-BLS) either lacks an efficient training method and incremental learning capability or suffers from poor performance. To this end, we propose a convolutional broad learning system (ConvBLS) based on the spherical K-means (SKM) algorithm and two-stage multi-scale (TSMS) feature fusion, which consists of the convolutional feature (CF) layer, convolutional enhancement (CE) layer, TSMS feature fusion layer, and output layer. First, unlike the current C-BLS, the simple yet efficient SKM algorithm is utilized to learn the weights of CF layers. Compared with random filters, the SKM algorithm makes the CF layer learn more comprehensive spatial features. Second, similar to the vanilla BLS, CE layers are established to expand the feature space. Third, the TSMS feature fusion layer is proposed to extract more effective multi-scale features through the integration of CF layers and CE layers. Thanks to the above design and the pseudo-inverse calculation of the output layer weights, our proposed ConvBLS method is unprecedentedly efficient and effective. Finally, the corresponding incremental learning algorithms are presented for rapid remodeling if the model deems to expand. Experiments and comparisons demonstrate the superiority of our method. Copyright © 2023, The Authors. All rights reserved.

关键词： Learning systems

来源：评论

学校读者我要写书评

暂无评论

A Distributed Inference System for Detecting Task-wise Single Trial Event-Related Potential in Stream of Satellite Images

arXiv

引用

arXiv 2023年

作者： Kim, Sung-Jin Kwak, Heon-Gyu Han, Hyeon-Taek Lee, Dae-Hyeok Jeong, Ji-Hoon Lee, Seong-Whan Dept. of Artificial Intelligence Korea University Seoul Korea Republic of Dept. of Brain and Cognitive Engineering Korea University Seoul Korea Republic of Dept. of Computer Science Chungbuk National University Cheongju Korea Republic of

brain-computer interface (BCI) has garnered the significant attention for their potential in various applications, with event-related potential (ERP) performing a considerable role in BCI systems. This paper introduces a novel Distributed Inference System tailored for detecting task-wise single-trial ERPs in a stream of satellite images. Unlike traditional methodologies that employ a single model for target detection, our system utilizes multiple models, each optimized for specific tasks, ensuring enhanced performance across varying image transition times and target onset times. Our experiments, conducted on four participants, employed two paradigms: the Normal paradigm and an AI paradigm with bounding boxes. Results indicate that our proposed system outperforms the conventional methods in both paradigms, achieving the highest Fβ scores. Furthermore, including bounding boxes in the AI paradigm significantly improved target recognition. This study underscores the potential of our Distributed Inference System in advancing the field of ERP detection in satellite image streams. © 2023, CC BY.

关键词： brain computer interface

来源：评论

学校读者我要写书评

暂无评论

Refined Risk Bounds for Unbounded Losses via Transductive Priors

arXiv

引用

arXiv 2024年

作者： Qian, Jian Rakhlin, Alexander Zhivotovskiy, Nikita Electrical Engineering and Computer Science Department Massachusetts Institute of Technology CambridgeMA02139 United States Department of Brain and Cognitive Sciences Massachusetts Institute of Technology CambridgeMA02139 United States Department of Statistics University of California Berkeley BerkeleyCA94720 United States

We revisit the sequential variants of linear regression with the squared loss, classification problems with hinge loss, and logistic regression, all characterized by unbounded losses in the setup where no assumptions are made on the magnitude of design vectors and the norm of the optimal vector of parameters. The key distinction from existing results lies in our assumption that the set of design vectors is known in advance (though their order is not), a setup sometimes referred to as transductive online learning. While this assumption might seem similar to fixed design regression or denoising, we demonstrate that the sequential nature of our algorithms allows us to convert our bounds into statistical ones with random design without making any additional assumptions about the distribution of the design vectors—an impossibility for standard denoising results. Our key tools are based on the exponential weights algorithm with carefully chosen transductive (design-dependent) priors, which exploit the full horizon of the design vectors, as well as additional aggregation tools that address the possibly unbounded norm of the vector of the optimal solution. Our classification regret bounds have a feature that is only attributed to bounded losses in the literature: they depend solely on the dimension of the parameter space and on the number of rounds, independent of the design vectors or the norm of the optimal solution. For linear regression with squared loss, we further extend our analysis to the sparse case, providing sparsity regret bounds that depend additionally only on the magnitude of the response variables. We argue that these improved bounds are specific to the transductive setting and unattainable in the worst-case sequential setup. Our algorithms, in several cases, have polynomial time approximations and reduce to sampling with respect to log-concave measures instead of aggregating over hard-to-construct Ε-covers of classes. Copyright © 2024, The Authors. All righ

关键词： Vectors

来源：评论

学校读者我要写书评

暂无评论

Deep learning framework for action prediction reveals multi-timescale locomotor control

arXiv

引用

arXiv 2025年

作者： Wang, Wei-Chen De Comite, Antoine Voloshina, Alexandra Daley, Monica Seethapathi, Nidhi Department of Electrical Engineering and Computer Science School of Engineering MIT CambridgeMA02139 United States McGovern Institute for Brain Research MIT CambridgeMA02139 United States Brain and Cognitive Sciences School of Science MIT CambridgeMA02139 United States Mechanical and Aerospace Engineering Henry Samueli School of Engineering UC Irvine IrvineCA92697 United States Ecology and Evolutionary Biology School of Biological Sciences UC Irvine IrvineCA92697 United States Biomedical Engineering Henry Samueli School of Engineering UC Irvine IrvineCA92697 United States

Modeling human movement in real-world tasks is a fundamental goal for motor control, biomechanics, and rehabilitation engineering. However, existing models of essential tasks like locomotion are not applicable across varying terrain, mechanical conditions, and sensory contexts. This is at least in part due to simplifying assumptions like linear and fixed timescales mappings between inputs and future actions, which may not be broadly applicable. Here, we develop a deep learning-based framework for action prediction, outperforming traditional models across multiple contexts (walking and running, treadmill and overground, varying terrains) and input modalities (multiple body states, visual gaze). We find that neural network architectures with flexible input history-dependence, like GRU and Transformer, and with architecture-dependent trial embeddings perform best overall. By quantifying the model’s predictions relative to an autoregressive baseline, we identify context- and modality-dependent timescales. These analyses reveal that there is greater reliance on fast-timescale predictions in complex terrain, gaze predicts future foot placement before body states, and the full-body state predictions precede those by center-of-mass states. This deep learning framework for human action prediction provides quantifiable insights into the control of real-world locomotion and can be extended to other actions, contexts, and populations. © 2025, CC BY-NC-SA.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

What did I sign? A study of the impenetrability of legalese in contracts 43

What did I sign? A study of the impenetrability of legalese ...

引用

43rd Annual Meeting of the cognitive science Society: Comparative Cognition: Animal Minds, CogSci 2021

作者： Martínez, Eric Mollica, Francis Liu, Yufei Podrug, Anita Gibson, Edward Brain and Cognitive Sciences MIT Cambridge United States School of Informatics University of Edinburgh Edinburgh United Kingdom Psychology Wellesley College Wellesley United States Electrical Engineering and Computer Science MIT Cambridge United States

Legal documents, in the form of terms of service agreements and other private contracts, are now an increasingly prevalent part of everyday life. While legal documents have long been acknowledged to be difficult to understand without training, it remains an open question whether the ever-increasing exposure to contracts might have mitigated this difficulty. Moreover, insofar as this difficulty has persisted, there remains no systematic analysis of which linguistic structures contribute most heavily to the processing difficulty of legal texts, nor whether this difficulty is heightened for those with less language experience. Here, we investigate these issues, and in a well-powered experiment find evidence that (a) both recall and comprehension of legal propositions in a contract are hindered by use of a legal register relative to plain-English translations;(b) certain linguistic structures, such as center-embedding, hinder recall to a greater degree than others, such as passive voice;and (c) language experience influences comprehension of legal propositions. Surprisingly, language experience did not influence recall, nor was there an interaction between legal register and language experience on recall or comprehension. These findings suggest that legal language poses heightened difficulties for those with less language experience–who tend to be of lower socioeconomic status and with diminished access to the justice system–and that eliminating complex features of legalese would benefit those of all reading levels. © cognitive science Society: Comparative Cognition: Animal Minds, CogSci *** rights reserved.

关键词： Contracts

来源：评论

学校读者我要写书评

暂无评论

Depression Severity Identification Based on Shallow 2d Self-Attention-Cnn Using Eeg Functional Connectivity Network

SSRN

引用

SSRN 2024年

作者： Zhou, Yihan Yu, Xiaokang Lin, Huiping Li, Rihui Liang, Jiuxing Shi, Xue Luo, Yuxi School of Biomedical Engineering Shenzhen Campus of Sun Yat-sen University Guangdong Shenzhen518100 China Center for Cognitive and Brain Sciences Department of Electrical and Computer Engineering University of Macau Macau S.A.R Taipa China Key Laboratory of Brain Cognition and Education Science Ministry of Education Institute for Brain Research and Rehabilitation Guangdong Key Laboratory of Mental Health and Cognitive Science South China Normal University China Department of Neurology Shenzhen People's Hospital The Second Clinical Medical College Jinan University The First Affiliated Hospital Southern University of Science and Technology Shenzhen China Key Laboratory of Sensing Technology and Biomedical Instruments of Guangdong Province Sun Yat-sen University Guangdong Shenzhen518100 China

Background and objectives: Depression inflicts significant harm on both society and family. Previous studies have indicated that the functional network of EEG signals worked well in recognizing major depression. This study aims to further identify depression severities and characterize their EEG functional network difference by designing a deep learning strategy and a corresponding visualization ***: Rest state EEG recordings from 30 healthy controls, 35 mild depressed and 26 severe depressed patients were included. Weighted phase lag indexes were computed across four frequency sub-bands to delineate the functional connectivity of EEG networks, serving as the input matrix. To better adapt volume conduction effects, a shallow CNN-based incorporated with 2D Self-Attention architecture was designed, enabling the model to capture information across diverse spans and scales within the functional connectivity (FC) matrix. Leveraging the Grad-CAM algorithm, the model highlighted crucial FCs and corresponding EEG pairs for classification. Finally, the changes of EEG FC network across depression severities were statistically analyzed and ***: An accuracy of 89.2% was achieved in tri-classification of 10-second EEG segments using 52 EEG channels, remaining high at 84.1% with 30 selected channels. Notably, the investigation revealed that significant FC changes from mild to severe depression did not exhibit a simple or monotonous ***: This research presented a directly measured methodology to identify depression severity, vital for informing prevention and therapeutic interventions. Furthermore, the findings shed light on the evolving patterns of brain function as depression progresses. The proposed deep learning model and channel selection algorithm offer potential applications beyond this study, promising broader utility in EEG-based research endeavors. © 2024, The Authors. All rights reserved.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

SAM+nnUNet: Deep-learning-based Head and Neck Tumor Segmentation on FDG PET

SAM+nnUNet: Deep-learning-based Head and Neck Tumor Segmenta...

引用

IEEE Symposium on Nuclear science (NSS/MIC)

作者： Z. Fang Z. Lu H. Liu Y. Liu G. S.P. Mok Department of Electrical and Computer Engineering Faculty of Science and Technology Biomedical Imaging Laboratory (BIG) University of Macau Macau China Department of Nuclear Medicine The Affiliated Hospital Southwest Medical University Luzhou China Center for Cognitive and Brain Sciences Institute of Collaborative Innovation University of Macau Macau China

ISBN: (数字)9798350388152

ISBN: (纸本)9798350388169

FDG PET is a useful tool for diagnosis, staging and treatment monitoring on Head and Neck (H&N) tumors. Segmentation is usually performed for the tumors while the deep learning technique is emerging as an efficient yet precise alternate to manual segmentation. The performance of the conventional CNN-based segmentation networks is limited by insufficient training data. The Segment Anything Model (SAM) has been released by Meta, initiating a paradigm shift to a “smart” interactive segmentation that provides competitive performance when properly prompted. In this study, we proposed a SAM+nnUNet for H&N tumor segmentation on FDG PET to enhance the original SAM, in which the prompting process can be time-consuming and require a high degree of expertise. The HECKTOR2022 PET/CT dataset was used, with 420 patient images for training, 50 cases for validation, and 54 cases for testing. The training dataset of PET was used to fine-tune the mask decoder of pre-trained SAM, while both PET and CT were used to train an nnUNet. The trained nnUNet was first used to perform a preliminary segmentation of the tumors on the testing dataset, which was used to generate bounding box prompts for the SAM input after visual inspection and manual adjustment if necessary. Thresholding segmentation based on 42% of maximum intensity and various U-Net variants are implemented for comparison. Dice Similarity Coefficient (DSC), Hausdorff Distance at 95th percentile (HD95), and SUVmean of segmented tumors compared to the ground truth were analyzed. Our results indicate that 5 out of 54 cases need manual adjustment for the nnUNet-based prompts for SAM. SAM+nnUNet performs best in all quantitative indices. We conclude that our proposed method is promising for H&N tumor segmentation on PET.

关键词： Training Image segmentation Visualization Head Training data Manuals Artificial neural networks

来源：评论

学校读者我要写书评

暂无评论

Leveraging Unstructured Statistical Knowledge in a Probabilistic Language of Thought 42

Leveraging Unstructured Statistical Knowledge in a Probabili...

引用

42nd Annual Meeting of the cognitive science Society: Developing a Mind: Learning in Humans, Animals, and Machines, CogSci 2020

作者： Lew, Alexander K. Tessler, Michael Henry Mansinghka, Vikash K. Tenenbaum, Joshua B. Department of Electrical Engineering and Computer Science MIT United States Department of Brain and Cognitive Sciences MIT United States

One hallmark of human reasoning is that we can bring to bear a diverse web of common-sense knowledge in any situation. The vastness of our knowledge poses a challenge for the practical implementation of reasoning systems as well as for our cognitive theories - how do people represent their commonsense knowledge? On the one hand, our best models of sophisticated reasoning are top-down, making use primarily of symbolically-encoded knowledge. On the other, much of our understanding of the statistical properties of our environment may arise in a bottom-up fashion, for example through associationist learning mechanisms. Indeed, recent advances in AI have enabled the development of billion-parameter language models that can scour for patterns in gigabytes of text from the web, picking up a surprising amount of common-sense knowledge along the way-but they fail to learn the structure of coherent reasoning. We propose combining these approaches, by em- bedding language-model-backed primitives into a state-of-the-art probabilistic programming language (PPL). On two open-ended reasoning tasks, we show that our PPL models with neural knowledge components characterize the distribution of human responses more accurately than the neural language models alone, raising interesting questions about how people might use language as an interface to common-sense knowledge, and suggesting that building probabilistic models with neural language-model components may be a promising approach for more human-like AI. © 2020 The Author(s)

关键词： Probability distributions

来源：评论

学校读者我要写书评

暂无评论

Deep and Dynamic Metabolic and Structural Imaging in Living Tissues

arXiv

引用

arXiv 2024年

作者： Liu, Kunzan Cao, Honghao Shashaty, Kasey Yu, Li-Yu Spitz, Sarah Pramotton, Francesca Michela Wan, Zhengpeng Kan, Ellen L. Tevonian, Erin N. Levy, Manuel Lendaro, Eva Kamm, Roger D. Griffith, Linda G. Wang, Fan Qiu, Tong You, Sixian Research Laboratory of Electronics Massachusetts Institute of Technology United States Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology United States Department of Biological Engineering Massachusetts Institute of Technology United States Department of Brain and Cognitive Sciences Massachusetts Institute of Technology United States Department of Mechanical Engineering Massachusetts Institute of Technology United States

Label-free imaging through two-photon autofluorescence (2PAF) of NAD(P)H allows for non-destructive and high-resolution visualization of cellular activities in living systems. However, its application to thick tissues and organoids has been restricted by its limited penetration depth within 300 µm, largely due to tissue scattering at the typical excitation wavelength (∼750 nm) required for NAD(P)H. Here, we demonstrate that the imaging depth for NAD(P)H can be extended to over 700 µm in living engineered human multicellular microtissues by adopting multimode fiber (MMF)-based low-repetition-rate high-peak-power three-photon (3P) excitation of NAD(P)H at 1100 nm. This is achieved by having over 0.5 MW peak power at the band of 1100±25 nm through adaptively modulating multimodal nonlinear pulse propagation with a compact fiber shaper. Moreover, the 8-fold increase in pulse energy at 1100 nm enables faster imaging of monocyte behaviors in the living multicellular models. These results represent a significant advance for deep and dynamic metabolic and structural imaging of intact living biosystems. The modular design (MMF with a slip-on fiber shaper) is anticipated to allow wide adoption of this methodology for demanding in vivo and in vitro imaging applications, including cancer research, autoimmune diseases, and tissue engineering. © 2024, CC BY.

关键词： Multimode fibers

来源：评论

学校读者我要写书评

暂无评论

A CONFORMER BASED ACOUSTIC MODEL FOR ROBUST AUTOMATIC SPEECH RECOGNITION

arXiv

引用

arXiv 2022年

作者： Yang, Yufeng Wang, Peidong Wang, DeLiang Department of Computer Science and Engineering The Ohio State University United States Microsoft Corporation United States Center for Cognitive and Brain Sciences The Ohio State University United States

This study addresses robust automatic speech recognition (ASR) by introducing a Conformer-based acoustic model. The proposed model builds on the wide residual bi-directional long short-term memory network (WRBN) with utterance-wise dropout and iterative speaker adaptation, but employs a Conformer encoder instead of the recurrent network. The Conformer encoder uses a convolution-augmented attention mechanism for acoustic modeling. The proposed system is evaluated on the monaural ASR task of the CHiME-4 corpus. Coupled with utterance-wise normalization and speaker adaptation, our model achieves 6.25% word error rate, which outperforms WRBN by 8.4% relatively. In addition, the proposed Conformer-based model is 18.3% smaller in model size and reduces total training time by 79.6%. Copyright © 2022, The Authors. All rights reserved.

关键词： Acoustic Modeling

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

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

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：