我们为梯度下降提供了收敛分析,以解决高斯分布中不可知的问题。与研究零偏差的设置的先前工作不同,我们考虑了当relu函数的偏见非零时更具挑战性的情况。我们的主要结果确定,从随机初始化开始,从多项式迭代梯度下降输出中,具有很高的概率,与最佳relu函数的误差相比,可以实现竞争错误保证。我们还提供有限的样本保证,这些技术将其推广到高斯以外的更广泛的边际分布。
translated by 谷歌翻译
在本文中,我们提出了一个自然的单个偏好(IP)稳定性的概念,该概念要求每个数据点平均更接近其自身集群中的点,而不是其他群集中的点。我们的概念可以从几个角度的动机,包括游戏理论和算法公平。我们研究了与我们提出的概念有关的几个问题。我们首先表明,确定给定数据集通常允许进行IP稳定的聚类通常是NP-HARD。结果,我们探索了在某些受限度量空间中查找IP稳定聚类的有效算法的设计。我们提出了一种poly Time算法,以在实际线路上找到满足精确IP稳定性的聚类,并有效地算法来找到针对树度量的IP稳定2聚类。我们还考虑放松稳定性约束,即,与其他任何集群相比,每个数据点都不应太远。在这种情况下,我们提供具有不同保证的多时间算法。我们在实际数据集上评估了一些算法和几种标准聚类方法。
translated by 谷歌翻译
我们研究在对抗性腐败下学习通用线性模型的问题。我们分析了一种经典的启发式,称为迭代修剪的最大似然估计量,该估计量已知在实践中有效地抵抗标签腐败。在标签腐败下,我们证明了这个简单的估计器在各种广泛的线性模型上实现了最小风险,包括高斯回归,泊松回归和二项式回归。最后,我们将估计器扩展到标签和协变量腐败的更具挑战性的设置,并在该环境中证明其稳健性和最佳性。
translated by 谷歌翻译
噪声对比度估计的最新研究表明,从经验上讲,从理论上讲,尽管在对比度损失中拥有更多的“负样本”,但最初在阈值中提高了下游分类的性能,但由于“碰撞覆盖“贸易”,它都会损害下游性能-离开。但是,对比度学习中固有的现象是如此吗?我们在一个简单的理论环境中显示,通过从基础潜在类采样(由Saunshi等人引入(ICML 2019)),产生正对,表明表示(人口)对比度损失的下游性能实际上确实确实确实如此。不会随着负样本的数量降低。一路上,我们在框架中给出了最佳表示形式的结构表征,以进行噪声对比估计。我们还为CIFAR-10和CIFAR-100数据集的理论结果提供了经验支持。
translated by 谷歌翻译
对抗性鲁棒性是各种现代机器学习应用中的关键财产。虽然它是最近几个理论研究的主题,但与对抗性稳健性有关的许多重要问题仍然是开放的。在这项工作中,我们研究了有关对抗对抗鲁棒性的贝叶斯最优性的根本问题。我们提供了一般的充分条件,可以保证贝叶斯最佳分类器的存在,以满足对抗性鲁棒性。我们的结果可以提供一种有用的工具,用于随后研究对抗性鲁棒性及其一致性的替代损失。这份稿件是“关于普通贝叶斯分类器的存在”在神经潮端中发表的延伸版本。原始纸张的结果不适用于一些非严格凸的规范。在这里,我们将结果扩展到所有可能的规范。
translated by 谷歌翻译
我们提出了简单的主动采样和重新重量策略,以优化最小最大公平性,可以应用于通过损耗最小化学习的任何分类或回归模型。我们的方法背后的关键直觉是在每个TIMESTEP中使用来自当前模型中最差的组的DataPoint,以更新模型。实施的易于实现和我们稳健的制定的一般性使其成为提高糟糕表现群体的模型性能的有吸引力的选择。对于凸起的学习问题,如线性或逻辑回归,我们提供了对我们的策略的细粒度分析,证明了其收敛速度对Min-Max Fair解决方案。
translated by 谷歌翻译
Language models have become increasingly popular in recent years for tasks like information retrieval. As use-cases become oriented toward specific domains, fine-tuning becomes default for standard performance. To fine-tune these models for specific tasks and datasets, it is necessary to carefully tune the model's hyperparameters and training techniques. In this paper, we present an in-depth analysis of the performance of four transformer-based language models on the task of biomedical information retrieval. The models we consider are DeepMind's RETRO (7B parameters), GPT-J (6B parameters), GPT-3 (175B parameters), and BLOOM (176B parameters). We compare their performance on the basis of relevance, accuracy, and interpretability, using a large corpus of 480000 research papers on protein structure/function prediction as our dataset. Our findings suggest that smaller models, with <10B parameters and fine-tuned on domain-specific datasets, tend to outperform larger language models on highly specific questions in terms of accuracy, relevancy, and interpretability by a significant margin (+50% on average). However, larger models do provide generally better results on broader prompts.
translated by 谷歌翻译
This work aims at showing that it is feasible and safe to use a swarm of Unmanned Aerial Vehicles (UAVs) indoors alongside humans. UAVs are increasingly being integrated under the Industry 4.0 framework. UAV swarms are primarily deployed outdoors in civil and military applications, but the opportunities for using them in manufacturing and supply chain management are immense. There is extensive research on UAV technology, e.g., localization, control, and computer vision, but less research on the practical application of UAVs in industry. UAV technology could improve data collection and monitoring, enhance decision-making in an Internet of Things framework and automate time-consuming and redundant tasks in the industry. However, there is a gap between the technological developments of UAVs and their integration into the supply chain. Therefore, this work focuses on automating the task of transporting packages utilizing a swarm of small UAVs operating alongside humans. MoCap system, ROS, and unity are used for localization, inter-process communication and visualization. Multiple experiments are performed with the UAVs in wander and swarm mode in a warehouse like environment.
translated by 谷歌翻译
Human Activity Recognition (HAR) using on-body devices identifies specific human actions in unconstrained environments. HAR is challenging due to the inter and intra-variance of human movements; moreover, annotated datasets from on-body devices are scarce. This problem is mainly due to the difficulty of data creation, i.e., recording, expensive annotation, and lack of standard definitions of human activities. Previous works demonstrated that transfer learning is a good strategy for addressing scenarios with scarce data. However, the scarcity of annotated on-body device datasets remains. This paper proposes using datasets intended for human-pose estimation as a source for transfer learning; specifically, it deploys sequences of annotated pixel coordinates of human joints from video datasets for HAR and human pose estimation. We pre-train a deep architecture on four benchmark video-based source datasets. Finally, an evaluation is carried out on three on-body device datasets improving HAR performance.
translated by 谷歌翻译
如今,瑜伽因现代生活方式的压力增加而受到全世界的关注,并且学习瑜伽有很多方法或资源。瑜伽一词意味着思想和身体之间的深厚联系。今天,有大量的医学和科学证据表明,我们大脑活动的基本面,我们的化学甚至可以通过练习不同的瑜伽系统来改变我们的化学。 Suryanamaskar,也被称为“向太阳致敬”,是一种瑜伽练习,结合了八种不同的形式和12个体式(4个Asana重复),专门介绍了印度太阳神Surya。 Suryanamaskar提供了许多健康益处,例如增强肌肉和帮助控制血糖水平。在这里,MediaPipe库用于分析Surya Namaskar的情况。高级软件可以实时检测到站立,因为人们在相机前表演了Surya Namaskar。班级分隔器将该表格识别为以下一项:pranamasana,hasta padasana,hasta uttanasana,ashwa -Sanchalan Asana,Ashtanga Namaskar,Dandasana或Bhujangasana和Svanasana。基于深度学习的技术(CNN)用于开发该模型,模型精度为98.68%,精度得分为0.75,以检测正确的瑜伽(Surya Namaskar)姿势。使用此方法,用户可以练习所需的姿势,并可以检查该人所做的姿势是否正确。它将有助于正确地做Surya Namaskar的所有不同姿势,并提高瑜伽从业者的效率。本文描述了将在模型中实现的整个框架。
translated by 谷歌翻译