paper
GitHub - Marsrocky/Awesome-WiFi-CSI-Sensing: A list of awesome papers and cool resources on WiFi CSI sensing.
视频介绍
Underline Science | Access Academic Videos & Scientific Conferences Online
paper review
Venues | OpenReview
paper review
Full article: Peer Review Questions & Answers: How? Part III: Writing the Reviewer Report (tandfonline.com)
Major comments should reflect the manuscript’s most concerning areas. These generally include critical points that the authors must address for the paper to be considered for publication (Citation9, Citation10, Citation12, Citation13). Revision and clarification of these major issues will be fundamental to understanding the manuscript. Minor issues, although important, are considered ...
HAR
activities“Human activities can be categorized into different types [5], including locomotion (e.g., walking, running, standing, and still), exercise (e.g., cycling and playing soccer), health related activities (e.g., falls, rehabilitation, and following routines), daily activities (e.g., shopping, using computer, sleeping, going to work, and attending a meeting), and so on” (Gu 等, 2018, p. 2085)
论文-method
solutions for data scarcity
using pretrained networks as feature extractors, (While transfer learning can mitigate data shortages, pre-trained networks are only useful when their inputs are similar to the data that they were trained for)
co training, a classifier that is already trained with an existing labeled dataset is used to classify the unlabeled data. (“Accumulated error is common in co-training, and the newly labeled data has a limited enhancement in classifier performance.” (Mohammadza ...
论文速读-综述
Continuous Human Action Recognition for Human-machine Interaction: A Review
这篇综述1介绍了连续人类活动识别(HAR)的概念,方法,应用,和挑战。连续HAR是指利用传感器或视频数据来检测和识别人类的行为和动作,并且能够在输入视频中分割出不同的动作片段。连续HAR对于需要实时人机交互的应用非常重要,如智能家居,健康监测,安防监控等。然而,连续HAR也面临着一些挑战,如数据的多样性,复杂性,噪声,不平衡等。
这篇综述回顾了近年来的相关文献,详细分析,解释,和比较了不同的动作分割方法,以及它们使用的特征提取和学习策略。这些方法主要分为两类:基于监督学习的方法和基于无监督学习的方法。基于监督学习的方法需要有标注的动作片段作为训练数据,通常使用深度神经网络,如卷积神经网络(CNN),循环神经网络(RNN),或者注意力机制来学习时空特征和动作边界。基于无监督学习的方法不需要有标注的动作片段,而是利用数据的内在结构,如变化点,子空间,或者聚类来发现动作边界。这篇综述还介绍了一些混合的方法,如半监督学习,弱监督学习,和自 ...
Generative AI-LLM
斯坦福论文《Generative Agents》用 AI 角色模拟人类行为,能带来哪些应用? - 知乎 (zhihu.com)microsoft/PromptCraft-Robotics: Community for applying LLMs to robotics and a robot simulator with ChatGPT integration (github.com)25 ChatGPT AIs Play A Game - So What Happened?_哔哩哔哩_bilibiliLC1332/Chinese-generative-agents: 斯坦福工作 Generative Agents的复现和翻译 An attempt to build a working, locally-running cheap version of Generative Agents: Interactive Simulacra of Human Behavior (github.com)
主流模型架构大模型面经——从prefix-decoder、casua ...