混合样品正则化(MSR),例如混合或cutmix,是一种强大的数据增强策略,可以推广卷积神经网络。先前的经验分析说明了MSR与传统的离线知识蒸馏(KD)之间的正交性能增长。更具体地说,可以通过MSR参与顺序蒸馏的训练阶段来增强学生网络。然而,MSR和在线知识蒸馏之间的相互作用,这是一个更强的蒸馏范式,在那里,一群同伴互相学习的合奏仍然没有探索。为了弥合差距,我们首次尝试将cutmix纳入在线蒸馏中,我们从经验上观察到了重大改进。在这个事实的鼓舞下,我们提出了一个更强大的MSR,专门用于在线蒸馏,称为Cut^nMix。此外,一个新颖的在线蒸馏框架是在切割^nmix上设计的,以通过功能水平相互学习和自我启动的老师来增强蒸馏。对CIFAR10和CIFAR100进行六个网络体系结构的全面评估表明,我们的方法可以始终超过最先进的蒸馏方法。
translated by 谷歌翻译
现实世界中的数据通常显示出不平衡的标签分布。有关数据不平衡的现有研究集中在单域设置上,即样本来自相同的数据分布。但是,自然数据可以起源于不同的领域,在一个领域中的少数族裔可以从其他域中具有丰富的实例。我们正式化了多域长尾识别(MDLT)的任务,该任务从多域不平衡数据中学习,解决了跨域的标签不平衡,域移动和不同标签分布,并将其推广到所有域级对。我们首先开发了域类的可传递性图,并表明这种可传递性决定了MDLT中学习的成功。然后,我们提出了Boda,这是一种理论上的学习策略,可以跟踪可转移性统计的上限,并确保跨域级分布之间的平衡对齐和校准。我们策划了基于广泛使用的多域数据集的五个MDLT基准测试,并将BODA与跨越不同学习策略的二十个算法进行比较。广泛而严格的实验验证了BODA的出色性能。此外,作为副产品,Boda建立了有关域泛化基准测试的新的最新最先进,强调了解决跨域数据不平衡的重要性,这对于改善概括至看不见的域可能至关重要。代码和数据可在以下网址获得:https://github.com/yyzharry/multi-domain-mmbalance。
translated by 谷歌翻译
现实世界数据往往展现出长期分布,重量级别不平衡,其中大多数课程可以主导培训过程并改变少数阶层的决策边界。最近,研究人员调查了监督对长尾识别的对比学习的潜力,并证明它提供了强大的性能增益。在本文中,我们表明,虽然监督对比学习可以有助于提高性能,但过去的基线通过不平衡数据分布引入的均匀性差。这种差的均匀性在来自特征空间中具有差的少数阶级的样品中表现出来。为了解决这个问题,我们提出了有针对性的监督对比学习(TSC),从而提高了极度上的特征分布的均匀性。 TSC首先生成一组均匀分布在极度上的目标。然后,在训练期间使不同类别的特征会聚到这些不同的和均匀分布的目标。这迫使所有类别,包括少数群体类别,以维持特征空间中的统一分布,改善了类边界,即使在存在长尾数据的情况下也能提供更好的泛化。多个数据集的实验表明,TSC在长尾识别任务上实现了最先进的性能。
translated by 谷歌翻译
在复杂环境中开发针对四足动物的强大视觉引导控制器,具有各种障碍,动力环境和不平坦的地形,这是非常具有挑战性的。尽管增强学习(RL)为敏捷的运动技能提供了有希望的范式,并在模拟中提供了视觉投入,但在现实世界中将RL政策部署仍然非常具有挑战性。我们的关键见解是,除了域间隙的差异,模拟和现实世界之间的视觉外观外,控制管道的延迟也是困难的主要原因。在本文中,我们建议在训练RL代理时解决此问题。具体而言,我们通过使用过去的观测值模拟真实硬件的延迟,并以随机时期进行采样,以进行本体感受和视觉。我们在没有任何预定义的控制器或参考运动的情况下训练RL策略在物理模拟器中以端到端的控制,并将其直接部署在野外运行的真实A1四倍的机器人上。我们在具有复杂地形和障碍的不同室外环境中评估我们的方法。我们证明机器人可以高速操纵,避免障碍物,并在基准方面显示出显着改善。我们的带有视频的项目页面位于https://mehooz.github.io/mmdr-wild/。
translated by 谷歌翻译
虽然对理解计算机视觉中的手对象交互进行了重大进展,但机器人执行复杂的灵巧操纵仍然非常具有挑战性。在本文中,我们提出了一种新的平台和管道DEXMV(来自视频的Dexerous操纵)以进行模仿学习。我们设计了一个平台:(i)具有多指机器人手和(ii)计算机视觉系统的复杂灵巧操纵任务的仿真系统,以记录进行相同任务的人类手的大规模示范。在我们的小说管道中,我们从视频中提取3D手和对象姿势,并提出了一种新颖的演示翻译方法,将人类运动转换为机器人示范。然后,我们将多个仿制学习算法与演示进行应用。我们表明,示威活动确实可以通过大幅度提高机器人学习,并解决独自增强学习无法解决的复杂任务。具有视频的项目页面:https://yzqin.github.io/dexmv
translated by 谷歌翻译
Wearable sensors for measuring head kinematics can be noisy due to imperfect interfaces with the body. Mouthguards are used to measure head kinematics during impacts in traumatic brain injury (TBI) studies, but deviations from reference kinematics can still occur due to potential looseness. In this study, deep learning is used to compensate for the imperfect interface and improve measurement accuracy. A set of one-dimensional convolutional neural network (1D-CNN) models was developed to denoise mouthguard kinematics measurements along three spatial axes of linear acceleration and angular velocity. The denoised kinematics had significantly reduced errors compared to reference kinematics, and reduced errors in brain injury criteria and tissue strain and strain rate calculated via finite element modeling. The 1D-CNN models were also tested on an on-field dataset of college football impacts and a post-mortem human subject dataset, with similar denoising effects observed. The models can be used to improve detection of head impacts and TBI risk evaluation, and potentially extended to other sensors measuring kinematics.
translated by 谷歌翻译
Few-shot relation extraction (FSRE) aims at recognizing unseen relations by learning with merely a handful of annotated instances. To generalize to new relations more effectively, this paper proposes a novel pipeline for the FSRE task based on queRy-information guided Attention and adaptive Prototype fuSion, namely RAPS. Specifically, RAPS first derives the relation prototype by the query-information guided attention module, which exploits rich interactive information between the support instances and the query instances, in order to obtain more accurate initial prototype representations. Then RAPS elaborately combines the derived initial prototype with the relation information by the adaptive prototype fusion mechanism to get the integrated prototype for both train and prediction. Experiments on the benchmark dataset FewRel 1.0 show a significant improvement of our method against state-of-the-art methods.
translated by 谷歌翻译
Recent 3D-based manipulation methods either directly predict the grasp pose using 3D neural networks, or solve the grasp pose using similar objects retrieved from shape databases. However, the former faces generalizability challenges when testing with new robot arms or unseen objects; and the latter assumes that similar objects exist in the databases. We hypothesize that recent 3D modeling methods provides a path towards building digital replica of the evaluation scene that affords physical simulation and supports robust manipulation algorithm learning. We propose to reconstruct high-quality meshes from real-world point clouds using state-of-the-art neural surface reconstruction method (the Real2Sim step). Because most simulators take meshes for fast simulation, the reconstructed meshes enable grasp pose labels generation without human efforts. The generated labels can train grasp network that performs robustly in the real evaluation scene (the Sim2Real step). In synthetic and real experiments, we show that the Real2Sim2Real pipeline performs better than baseline grasp networks trained with a large dataset and a grasp sampling method with retrieval-based reconstruction. The benefit of the Real2Sim2Real pipeline comes from 1) decoupling scene modeling and grasp sampling into sub-problems, and 2) both sub-problems can be solved with sufficiently high quality using recent 3D learning algorithms and mesh-based physical simulation techniques.
translated by 谷歌翻译
Sleep stage recognition is crucial for assessing sleep and diagnosing chronic diseases. Deep learning models, such as Convolutional Neural Networks and Recurrent Neural Networks, are trained using grid data as input, making them not capable of learning relationships in non-Euclidean spaces. Graph-based deep models have been developed to address this issue when investigating the external relationship of electrode signals across different brain regions. However, the models cannot solve problems related to the internal relationships between segments of electrode signals within a specific brain region. In this study, we propose a Pearson correlation-based graph attention network, called PearNet, as a solution to this problem. Graph nodes are generated based on the spatial-temporal features extracted by a hierarchical feature extraction method, and then the graph structure is learned adaptively to build node connections. Based on our experiments on the Sleep-EDF-20 and Sleep-EDF-78 datasets, PearNet performs better than the state-of-the-art baselines.
translated by 谷歌翻译
在各种机器学习问题中,包括转移,多任务,连续和元学习在内,衡量不同任务之间的相似性至关重要。最新的测量任务相似性的方法依赖于体系结构:1)依靠预训练的模型,或2)在任务上进行培训网络,并将正向转移用作任务相似性的代理。在本文中,我们利用了最佳运输理论,并定义了一个新颖的任务嵌入监督分类,该分类是模型的,无训练的,并且能够处理(部分)脱节标签集。简而言之,给定带有地面标签的数据集,我们通过多维缩放和串联数据集样品进行嵌入标签,并具有相应的标签嵌入。然后,我们将两个数据集之间的距离定义为其更新样品之间的2-Wasserstein距离。最后,我们利用2-wasserstein嵌入框架将任务嵌入到矢量空间中,在该空间中,嵌入点之间的欧几里得距离近似于任务之间提出的2-wasserstein距离。我们表明,与最佳传输数据集距离(OTDD)等相关方法相比,所提出的嵌入导致任务的比较显着更快。此外,我们通过各种数值实验证明了我们提出的嵌入的有效性,并显示了我们所提出的距离与任务之间的前进和向后转移之间的统计学意义相关性。
translated by 谷歌翻译