检索结果-内蒙古大学图书馆

Comment on Martínez-Delgado et al. Using Absorption Models for Insulin and Carbohydrates and Deep Leaning to Improve Glucose Level Predictions. Sensors 2021, 21, 5273

学校读者我要写书评

暂无评论

Sensors 2024年第13期24卷 4361页

作者： Misplon, Josiah Z. R. Saini, Varun Sloves, Brianna P. Meerts, Sarah H. Musicant, David R. Department of Computer Science Carleton College NorthfieldMN55057 United States Population Health Epic Systems VeronaWI53593 United States Neuroscience Program and Department of Psychology Carleton College NorthfieldMN55057 United States

The paper "Using Absorption Models for Insulin and Carbohydrates and Deep Leaning to Improve Glucose Level Predictions" (Sensors 2021, 21, 5273) proposes a novel approach to predicting blood glucose levels for people with type 1 diabetes mellitus (T1DM). By building exponential models from raw carbohydrate and insulin data to simulate the absorption in the body, the authors reported a reduction in their model’s root-mean-square error (RMSE) from (Formula presented.) mg/dL (raw) to (Formula presented.) mg/dL (exponential) when predicting blood glucose levels one hour into the future. In this comment, we demonstrate that the experimental techniques used in that paper are flawed, which invalidates its results and conclusions. Specifically, after reviewing the authors’ code, we found that the model validation scheme was malformed, namely, the training and test data from the same time intervals were mixed. This means that the reported RMSE numbers in the referenced paper did not accurately measure the predictive capabilities of the approaches that were presented. We repaired the measurement technique by appropriately isolating the training and test data, and we discovered that their models actually performed dramatically worse than was reported in the paper. In fact, the models presented in the that paper do not appear to perform any better than a naive model that predicts future glucose levels to be the same as the current ones. © 2024 by the authors.

关键词： Forecasting

Position: Bayesian Deep Learning is Needed in the Age of Large-Scale AI 41

学校读者我要写书评

暂无评论

Position: Bayesian Deep Learning is Needed in the Age of Lar...

41st International Conference on Machine Learning, ICML 2024

作者： Papamarkou, Theodore Skoularidou, Maria Palla, Konstantina Aitchison, Laurence Arbel, Julyan Dunson, David Filippone, Maurizio Fortuin, Vincent Hennig, Philipp Hernández-Lobato, José Miguel Hubin, Aliaksandr Immer, Alexander Karaletsos, Theofanis Khan, Mohammad Emtiyaz Kristiadi, Agustinus Li, Yingzhen Mandt, Stephan Nemeth, Christopher Osborne, Michael A. Rudner, Tim G.J. Rügamer, David Teh, Yee Whye Welling, Max Wilson, Andrew Gordon Zhang, Ruqi Department of Mathematics The University of Manchester Manchester United Kingdom Eric and Wendy Schmidt Center Broad Institute of MIT and Harvard Cambridge United States Spotify London United Kingdom Computational Neuroscience Unit University of Bristol Bristol United Kingdom Centre Inria de l'Université Grenoble Alpes Grenoble France Department of Statistical Science Duke University United States Statistics Program KAUST Saudi Arabia Helmholtz AI Munich Germany Department of Computer Science Technical University of Munich Munich Germany Munich Center for Machine Learning Munich Germany Tübingen AI Center University of Tübingen Tübingen Germany Department of Engineering University of Cambridge Cambridge United Kingdom Department of Mathematics University of Oslo Oslo Norway Bioinformatics and Applied Statistics Norwegian University of Life Sciences Ås Norway Department of Computer Science ETH Zurich Switzerland Chan Zuckerberg Initiative CA United States Center for Advanced Intelligence Project RIKEN Tokyo Japan Vector Institute Toronto Canada Department of Computing Imperial College London London United Kingdom Department of Computer Science UC Irvine Irvine United States Department of Mathematics and Statistics Lancaster University Lancaster United Kingdom Department of Engineering Science University of Oxford Oxford United Kingdom Center for Data Science New York University New York United States Department of Statistics LMU Munich Munich Germany DeepMind London United Kingdom Department of Statistics University of Oxford Oxford United Kingdom Informatics Institute University of Amsterdam Amsterdam Netherlands Courant Institute of Mathematical Sciences Center for Data Science Computer Science Department New York University New York United States Department of Computer Science Purdue University West Lafayette United States

In the current landscape of deep learning research, there is a predominant emphasis on achieving high predictive accuracy in supervised tasks involving large image and language datasets. However, a broader perspective reveals a multitude of overlooked metrics, tasks, and data types, such as uncertainty, active and continual learning, and scientific data, that demand attention. Bayesian deep learning (BDL) constitutes a promising avenue, offering advantages across these diverse settings. This paper posits that BDL can elevate the capabilities of deep learning. It revisits the strengths of BDL, acknowledges existing challenges, and highlights some exciting research avenues aimed at addressing these obstacles. Looking ahead, the discussion focuses on possible ways to combine large-scale foundation models with BDL to unlock their full potential. Copyright 2024 by the author(s)

关键词： Contrastive Learning

A large language model-assisted education tool to provide feedback on open-ended responses

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Matelsky, Jordan K. Parodi, Felipe Liu, Tony Lange, Richard D. Kording, Konrad P. Department of Bioengineering University of Pennsylvania United States Research & Exploratory Development Department Johns Hopkins University Applied Physics Laboratory United States Department of Neuroscience University of Pennsylvania United States Department of Computer Science University of Pennsylvania United States Department of Computer Science Rochester Institute of Technology United States CIFAR LMB Program

Open-ended questions are a favored tool among instructors for assessing student understanding and encouraging critical exploration of course material. Providing feedback for such responses is a time-consuming task that can lead to overwhelmed instructors and decreased feedback quality. Many instructors resort to simpler question formats, like multiple-choice questions, which provide immediate feedback but at the expense of personalized and insightful comments. Here, we present a tool that uses large language models (LLMs), guided by instructor-defined criteria, to automate responses to open-ended questions. Our tool delivers rapid personalized feedback, enabling students to quickly test their knowledge and identify areas for improvement. We provide open-source reference implementations both as a web application and as a Jupyter Notebook widget that can be used with instructional coding or math notebooks. With instructor guidance, LLMs hold promise to enhance student learning outcomes and elevate instructional methodologies. © 2023, CC BY-NC-SA.

关键词： Students

3D-to-2D-to-3D Conscious Learning

学校读者我要写书评

暂无评论

3D-to-2D-to-3D Conscious Learning

2022 IEEE International Conference on Consumer Electronics, ICCE 2022

作者： Weng, Juyang Department of Computer Science and Engineering Michigan State University East LansingMI48824 United States Cognitive Science Program Michigan State University East LansingMI48824 United States Neuroscience Program Michigan State University East LansingMI48824 United States Brain-Mind Institute OkemosMI48864 United States GENISAMA OkemosMI48864 United States

ISBN: (纸本)9781665441544

This is a theoretical paper on conscious learning for thoughts and creativity through general-purpose and autonomous imitation of demonstrations. This conscious learning is end-to-end (3D-to-2D-to-3D) and free from annotations of 2D images and 2D motor images (e.g., a bounding box to be attended to). The conscious learning algorithm directly takes that of the Developmental Networks that has been previously published extensively with rich experimental results. Apparently, humans and animals do this type of fully automated learning daily, but it is unclear a robot can do the same. Recently, [1], [2] presented a theory of conscious learning rooted in emergent universal Turing machines. It appeared to be the first algorithmic level theory of holistic consciousness, other than many papers in the literature about piecemeal consciousness. However, [1], [2] proved only conscious learning in motor-imposed training mode, namely 3D-to-2D taught by 2D motor impositions, free from 2D annotations. This paper fills the challenging gap in [1], [2] so the conscious learning is 3D-to-2D-to-3D (end-to-end) without motor-impositions or computing 'inverse kinematics'. This is a major departure from traditional AI-handcrafting symbolic labels that tend to be brittle (e.g., for driverless cars) and then 'spoon-feeding' pre-collected 'big data'. Autonomous imitations drastically reduce the teaching complexity compared to pre-collected 'big data', especially because no annotations of training data are needed. Furthermore, conscious learning allows creativity beyond what is taught. This work is directly related to consumer electronics because it requires large-scale on-the-fly brainoid chips in future wearable robots/devices for consumers. © 2022 IEEE.

关键词： Big data

Simulated prosthetic vision identifies checkerboard as an effective raster pattern for retinal implants

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Kasowski, Justin M. Beyeler, Michael Interdepartmental Graduate Program in Dynamical Neuroscience University of California Santa BarbaraCA93106 United States Department of Computer Science University of California Santa BarbaraCA93106 United States Department of Psychological & Brain Sciences University of California Santa BarbaraCA93106 United States

Objective: This study systematically evaluates the impact of raster patterns-specific spatial arrangements of sequential electrode activation-on performance and perceived difficulty in simulated prosthetic vision (SPV). By addressing a critical gap in the literature, we aimed to identify patterns that optimize functional vision in retinal implants. Approach: Sighted participants completed letter recognition and motion discrimination tasks under four raster patterns (horizontal, vertical, checkerboard, and random) using an immersive SPV system. The simulations employed psychophysically validated models of electrode activation, phosphene appearance, and temporal dynamics, ensuring realistic representation of prosthetic vision. Performance accuracy and self-reported difficulty were analyzed to assess the effects of raster patterning. Main Results: The checkerboard pattern consistently outperformed other raster patterns, yielding significantly higher accuracy and lower difficulty ratings across both tasks. The horizontal and vertical patterns introduced biases aligned with apparent motion artifacts, while the checkerboard minimized such effects. Random patterns resulted in the lowest performance, underscoring the importance of structured activation. Significance: These findings provide the first systematic evaluation of raster patterns, suggesting that the checkerboard configuration may enhance usability and perceptual clarity in retinal implants. While based on simulated environments, this study establishes a foundation for optimizing spatial and temporal electrode activation strategies in next-generation devices. © 2025, CC BY.

关键词： Prosthetics

Biomimetic Frontend for Differentiable Audio Processing

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Famularo, Ruolan Leslie Zotkin, Dmitry N. Shamma, Shihab A. Duraiswami, Ramani Perceptual Interfaces and Reality Lab. Computer Science & UMIACS University of Maryland College Park United States Department of Linguistics University of Maryland College Park United States Program in Neuroscience & Cognitive Science University of Maryland College Park United States Institute of Systems Research & Electrical and Computer Engineering University of Maryland College Park United States

While models in audio and speech processing are becoming deeper and more end-to-end, they as a consequence need expensive training on large data, and are often brittle. We build on a classical model of human hearing and make it differentiable, so that we can combine traditional explainable biomimetic signal processing approaches with deep-learning frameworks. This allows us to arrive at an expressive and explainable model that is easily trained on modest amounts of data. We apply this model to audio processing tasks, including classification and enhancement. Results show that our differentiable model surpasses black-box approaches in terms of computational efficiency and robustness, even with little training data. We also discuss other potential applications. © 2024, CC BY-NC-ND.

关键词： Deep learning

The use of chemogenetic actuator ligands in nonhuman primate DREADDs-fMRI

学校读者我要写书评

暂无评论

Current Research in Neurobiology 2023年 4卷 100072页

作者： Cushnie, Adriana K. Bullock, Daniel N. Manea, Ana M.G. Tang, Wei Zimmermann, Jan Heilbronner, Sarah R. Department of Neuroscience University of Minnesota Minneapolis 55455 MN United States Department of Computer Science Program in Neuroscience Indiana University Bloomington Bloomington 47408 IN United States Center for Magnetic Resonance Research University of Minnesota Minneapolis 55455 MN United States

Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) are engineered receptors that allow for genetically targeted, reversible manipulation of cellular activity via systemic drug administration. DREADD induced manipulations are initiated via the binding of an actuator ligand. Therefore, the use of DREADDs is contingent on the availability of actuator ligands. Actuator ligands low-dose clozapine (CLZ) and deschloroclozapine (DCZ) are highly selective for DREADDs, and, upon binding, induce physiological and behavioral changes in rodents and nonhuman primates (NHPs). Despite this reported specificity, both CLZ and DCZ have partial affinity for a variety of endogenous receptors and can induce dose-specific changes even in naïve animals. As such, this study aimed to examine the effects of CLZ and DCZ on resting-state functional connectivity (rs-FC) and intrinsic neural timescales (INTs) in naïve NHPs. In doing so, we evaluated whether CLZ and DCZ – in the absence of DREADDs – are inert by examining these ligands’ effects on the intrinsic functional properties of the brain. Low-dose DCZ did not induce consistent changes in rs-FC or INTs prior to the expression of DREADDs;however, a high dose resulted in subject-specific changes in rs-FC and INTs. In contrast, CLZ administration induced consistent changes in rs-FC and INTs prior to DREADD expression in our subjects. Our results caution against the use of CLZ by explicitly demonstrating the impact of off-target effects that can confound experimental results. Altogether, these data endorse the use of low dose DCZ for future DREADD-based experiments. © 2023

关键词： DREADDs Designer Receptors Exclusively Activated by Designer Drugs rs-fMRI Resting-State Functional Magnetic Resonance Imaging CLZ Clozapine DCZ Deschloroclozapine rs-FC Resting-State Functional Connectivity INTs Intrinsic Neural Timescales NHP Nonhuman Primate

Modeling Intensification for Sign Language Generation: A Computational Approach

学校读者我要写书评

暂无评论

arXiv 2022年

作者： İnan, Mert Zhong, Yang Hassan, Sabit Quandt, Lorna Alikhani, Malihe Computer Science Department School of Computing and Information University of Pittsburgh Pittsburgh United States Educational Neuroscience Program Gallaudet University WashingtonDC United States

End-to-end sign language generation models do not accurately represent the prosody in sign language. A lack of temporal and spatial variations leads to poor-quality generated presentations that confuse human interpreters. In this paper, we aim to improve the prosody in generated sign languages by modeling intensification in a data-driven manner. We present different strategies grounded in linguistics of sign language that inform how intensity modifiers can be represented in gloss annotations. To employ our strategies, we first annotate a subset of the benchmark PHOENIX-14T, a German Sign Language dataset, with different levels of intensification. We then use a supervised intensity tagger to extend the annotated dataset and obtain labels for the remaining portion of it. This enhanced dataset is then used to train state-of-the-art transformer models for sign language generation. We find that our efforts in intensification modeling yield better results when evaluated with automatic metrics. Human evaluation also indicates a higher preference of the videos generated using our model. © 2022, CC BY.

关键词： Modeling languages

Flexible Dual-Sided Polymer Microelectrode Array for Neurophysiological Recordings from an Insect Brain

学校读者我要写书评

暂无评论

Flexible Dual-Sided Polymer Microelectrode Array for Neuroph...

IEEE International Conference on Micro Electro Mechanical Systems

作者： Xiang Liu Simon Sanchez Yan Gong Zebin Jiang Trevor Stevens Hasan Banan Debajit Saha Wen Li Department of Physiology Neuroscience Program Michigan State University East Lansing MI USA Institute for Quantitative Health Science and Engineering East Lansing MI USA Department of Biomedical Engineering Michigan State University East Lansing MI USA Department of Electrical and Computer Engineering Michigan State University East Lansing MI USA

Despite advances in modern microelectrode technologies, designing high-density flexible microelectrode arrays (MEAs) for small insect brains remains challenging, especially when aiming for dense microelectrode configurations on flexible substrates in restricted areas (e.g., 250×100 μm 2 ). To overcome these challenges, we developed a novel dual-sided design that doubles the channel count and density of MEAs within the same spatial footprint in comparison with the single-sided counterparts. A unique folding-annealing method is introduced to craft Parylene-based flexible dual-sided density MEAs. These MEAs, with overall dimensions of 20 μm in thickness and 120-230 μm in width, are specifically tailored for the locust antenna lobe, which is the central odor processing circuitry in insects. The efficacy of the dual-sided MEA for neurophysiological sensing of odors was validated in vivo using a neural circuitry (antennal lobe) of the locust brain.

关键词：