本文在课堂增量学习中使用视觉变压器(VIT)研究。令人惊讶的是,天真地应用Vit替代卷积神经网络(CNNS)导致性能下降。我们的分析揭示了三个天然使用VIT的问题:(a)vit在课程中较小时具有非常缓慢的会聚,(b)在比CNN的模型中观察到新类的更多偏差,并且(c)适当的学习率Vit太低,无法学习良好的分类器。基于此分析,我们展示了这些问题可以简单地通过使用现有技术来解决:使用卷积杆,平衡FineTuning来纠正偏置,以及分类器的更高学习率。我们的简单解决方案名为Vitil(Vit用于增量学习),为所有三类增量学习设置实现了全新的最先进的保证金,为研究界提供了强大的基线。例如,在ImageNet-1000上,我们的体内体达到69.20%的前1个精度为500个初始类别的15个初始类别,5个增量步骤(每次100个新类),表现优于leulir + dde ​​1.69%。对于10个增量步骤(100个新课程)的更具挑战性的协议,我们的方法优于PODNet 7.27%(65.13%与57.86%)。
translated by 谷歌翻译
Modern machine learning suffers from catastrophic forgetting when learning new classes incrementally. The performance dramatically degrades due to the missing data of old classes. Incremental learning methods have been proposed to retain the knowledge acquired from the old classes, by using knowledge distilling and keeping a few exemplars from the old classes. However, these methods struggle to scale up to a large number of classes. We believe this is because of the combination of two factors: (a) the data imbalance between the old and new classes, and (b) the increasing number of visually similar classes. Distinguishing between an increasing number of visually similar classes is particularly challenging, when the training data is unbalanced. We propose a simple and effective method to address this data imbalance issue. We found that the last fully connected layer has a strong bias towards the new classes, and this bias can be corrected by a linear model. With two bias parameters, our method performs remarkably well on two large datasets: ImageNet (1000 classes) and MS-Celeb-1M (10000 classes), outperforming the state-of-the-art algorithms by 11.1% and 13.2% respectively.
translated by 谷歌翻译
在课堂增量学习(CIL)设置中,在每个学习阶段将类别组引入模型。目的是学习到目前为止观察到的所有类别的统一模型表现。鉴于视觉变压器(VIT)在常规分类设置中的最新流行,一个有趣的问题是研究其持续学习行为。在这项工作中,我们为CIL开发了一个伪造的双蒸馏变压器,称为$ \ textrm {d}^3 \ textrm {前} $。提出的模型利用混合嵌套的VIT设计,以确保数据效率和可扩展性对小数据集和大数据集。与最近的基于VIT的CIL方法相反,我们的$ \ textrm {d}^3 \ textrm {前} $在学习新任务并仍然适用于大量增量任务时不会动态扩展其体系结构。 $ \ textrm {d}^3 \ textrm {oft} $的CIL行为的改善归功于VIT设计的两个基本变化。首先,我们将增量学习视为一个长尾分类问题,其中大多数新课程的大多数样本都超过了可用于旧课程的有限范例。为了避免对少数族裔的偏见,我们建议动态调整逻辑,以强调保留与旧任务相关的表示形式。其次,我们建议在学习跨任务进行时保留空间注意图的配置。这有助于减少灾难性遗忘,通过限制模型以将注意力保留到最歧视区域上。 $ \ textrm {d}^3 \ textrm {以前} $在CIFAR-100,MNIST,SVHN和Imagenet数据集的增量版本上获得了有利的结果。
translated by 谷歌翻译
Conventionally, deep neural networks are trained offline, relying on a large dataset prepared in advance. This paradigm is often challenged in real-world applications, e.g. online services that involve continuous streams of incoming data. Recently, incremental learning receives increasing attention, and is considered as a promising solution to the practical challenges mentioned above. However, it has been observed that incremental learning is subject to a fundamental difficulty -catastrophic forgetting, namely adapting a model to new data often results in severe performance degradation on previous tasks or classes. Our study reveals that the imbalance between previous and new data is a crucial cause to this problem. In this work, we develop a new framework for incrementally learning a unified classifier, i.e. a classifier that treats both old and new classes uniformly. Specifically, we incorporate three components, cosine normalization, less-forget constraint, and inter-class separation, to mitigate the adverse effects of the imbalance. Experiments show that the proposed method can effectively rebalance the training process, thus obtaining superior performance compared to the existing methods. On CIFAR-100 and ImageNet, our method can reduce the classification errors by more than 6% and 13% respectively, under the incremental setting of 10 phases.
translated by 谷歌翻译
Although deep learning approaches have stood out in recent years due to their state-of-the-art results, they continue to suffer from catastrophic forgetting, a dramatic decrease in overall performance when training with new classes added incrementally. This is due to current neural network architectures requiring the entire dataset, consisting of all the samples from the old as well as the new classes, to update the model-a requirement that becomes easily unsustainable as the number of classes grows. We address this issue with our approach to learn deep neural networks incrementally, using new data and only a small exemplar set corresponding to samples from the old classes. This is based on a loss composed of a distillation measure to retain the knowledge acquired from the old classes, and a cross-entropy loss to learn the new classes. Our incremental training is achieved while keeping the entire framework end-to-end, i.e., learning the data representation and the classifier jointly, unlike recent methods with no such guarantees. We evaluate our method extensively on the CIFAR-100 and Im-ageNet (ILSVRC 2012) image classification datasets, and show state-of-the-art performance.
translated by 谷歌翻译
The dynamic expansion architecture is becoming popular in class incremental learning, mainly due to its advantages in alleviating catastrophic forgetting. However, task confusion is not well assessed within this framework, e.g., the discrepancy between classes of different tasks is not well learned (i.e., inter-task confusion, ITC), and certain priority is still given to the latest class batch (i.e., old-new confusion, ONC). We empirically validate the side effects of the two types of confusion. Meanwhile, a novel solution called Task Correlated Incremental Learning (TCIL) is proposed to encourage discriminative and fair feature utilization across tasks. TCIL performs a multi-level knowledge distillation to propagate knowledge learned from old tasks to the new one. It establishes information flow paths at both feature and logit levels, enabling the learning to be aware of old classes. Besides, attention mechanism and classifier re-scoring are applied to generate more fair classification scores. We conduct extensive experiments on CIFAR100 and ImageNet100 datasets. The results demonstrate that TCIL consistently achieves state-of-the-art accuracy. It mitigates both ITC and ONC, while showing advantages in battle with catastrophic forgetting even no rehearsal memory is reserved.
translated by 谷歌翻译
在这个不断变化的世界中,必须不断学习新概念的能力。但是,深层神经网络在学习新类别时会遭受灾难性的遗忘。已经提出了许多减轻这种现象的作品,而其中大多数要么属于稳定性困境,要么陷入了过多的计算或储存开销。受到梯度增强算法的启发,以逐渐适应目标模型和上一个合奏模型之间的残差,我们提出了一种新颖的两阶段学习范式寄养物,使该模型能够适应新的类别。具体而言,我们首先动态扩展新模块,以适合原始模型的目标和输出之间的残差。接下来,我们通过有效的蒸馏策略删除冗余参数和特征尺寸,以维护单个骨干模型。我们在不同的设置下验证CIFAR-100和Imagenet-100/1000的方法寄养。实验结果表明,我们的方法实现了最先进的性能。代码可在以下网址获得:https://github.com/g-u-n/eccv22-foster。
translated by 谷歌翻译
深度学习模型在逐步学习新任务时遭受灾难性遗忘。已经提出了增量学习,以保留旧课程的知识,同时学习识别新课程。一种典型的方法是使用一些示例来避免忘记旧知识。在这种情况下,旧类和新课之间的数据失衡是导致模型性能下降的关键问题。由于数据不平衡,已经设计了几种策略来纠正新类别的偏见。但是,他们在很大程度上依赖于新旧阶层之间偏见关系的假设。因此,它们不适合复杂的现实世界应用。在这项研究中,我们提出了一种假设不足的方法,即多粒性重新平衡(MGRB),以解决此问题。重新平衡方法用于减轻数据不平衡的影响;但是,我们从经验上发现,他们将拟合新的课程。为此,我们进一步设计了一个新颖的多晶正式化项,该项使模型还可以考虑除了重新平衡数据之外的类别的相关性。类层次结构首先是通过将语义或视觉上类似类分组来构建的。然后,多粒性正则化将单热标签向量转换为连续的标签分布,这反映了基于构造的类层次结构的目标类别和其他类之间的关系。因此,该模型可以学习类间的关系信息,这有助于增强新旧课程的学习。公共数据集和现实世界中的故障诊断数据集的实验结果验证了所提出的方法的有效性。
translated by 谷歌翻译
当在新的类或新任务上逐步训练时,深度神经网络易于灾难性遗忘,因为对新数据的适应导致旧课程和任务的性能急剧下降。通过使用小记忆进行排练和知识蒸馏,已证明最近的方法可有效缓解灾难性的遗忘。然而,由于内存的尺寸有限,旧的和新类可用的数据量之间的大不平衡仍然存在,这导致模型的整体精度恶化。为了解决这个问题,我们建议使用平衡的软制跨熵损失,并表明它可以与进出的方法相结合,以便在某些情况下降低培训过程的计算成本,以提高其性能。对竞争的想象,Subimagenet和CiFar100数据集的实验显示了最艺术态度的结果。
translated by 谷歌翻译
我们研究了类新型小说类发现的新任务(class-incd),该任务是指在未标记的数据集中发现新型类别的问题,该问题通过利用已在包含脱节的标签数据集上训练的预训练的模型,该模型已受过培训但是相关类别。除了发现新颖的课程外,我们还旨在维护模型识别先前看到的基本类别的能力。受到基于彩排的增量学习方法的启发,在本文中,我们提出了一种新颖的方法,以防止通过共同利用基类功能原型和特征级知识蒸馏来忘记对基础类的过去信息。我们还提出了一种自我训练的聚类策略,该策略同时将新颖的类别簇簇,并为基础和新颖类培训共同分类器。这使得我们的方法能够在课堂内设置中运行。我们的实验以三个共同的基准进行,表明我们的方法显着优于最先进的方法。代码可从https://github.com/oatmealliu/class-incd获得
translated by 谷歌翻译
我们考虑了类增量学习(CIL)问题,其中学习代理人通过逐步到达的培训数据批次不断学习新课程,并旨在在迄今为止所学的所有课程中很好地预测。问题的主要挑战是灾难性的遗忘,对于基于典范的示例性记忆方法,通常众所周知,遗忘通常是由于分类评分偏差引起的,该分类得分偏差是由于新类和新类之间的数据失衡而注射的旧课(在示例记忆中)。尽管已经提出了几种方法来通过一些其他后处理(例如,得分重新缩放或平衡的微调)来纠正这种分数偏见,但没有对这种偏见的根本原因进行系统分析。为此,我们分析了通过组合所有旧类和新类的输出得分来计算SoftMax概率的主要原因。然后,我们提出了一种新方法,称为分离的软磁性学习(SS-IL),该方法由分离的SoftMax(SS)输出层组成,结合了任务知识蒸馏(TKD)来解决此类偏见。在几个大规模CIL基准数据集的广泛实验结果中,我们通过在没有任何其他后处理的情况下获得更加平衡的预测分数来表明我们的SS-IL实现了强大的最新准确性。
translated by 谷歌翻译
深入学习模型遭受较旧阶段中课程的灾难性遗忘,因为它们在类增量学习设置中新阶段所引入的课程中受过培训。在这项工作中,我们表明灾难性忘记对模型预测的影响随着相同图像的方向的变化而变化,这是一种新的发现。基于此,我们提出了一种新的数据集合方法,该方法结合了图像的不同取向的预测,以帮助模型保留关于先前所见的类别的进一步信息,从而减少忘记模型预测的效果。但是,如果使用传统技术训练,我们无法直接使用数据集合方法。因此,我们还提出了一种新的双重增量学习框架,涉及共同培训网络,其中包括两个增量学习目标,即类渐进式学习目标以及我们提出的数据增量学习目标。在双增量学习框架中,每个图像属于两个类,即图像类(用于类增量学习)和方向类(用于数据增量学习)。在Class-Incremental学习中,每个新阶段都会引入一组新的类,并且模型无法从较旧阶段访问完整的培训数据。在我们提出的数据增量学习中,方向类在所有阶段保持相同,并且在类 - 增量学习中的新阶段引入的数据充当了这些方向类的新培训数据。我们经验证明双增量学习框架对数据集合方法至关重要。我们将拟议的课程逐步增量学习方法应用拟议方法,并经验表明我们的框架显着提高了这些方法的性能。
translated by 谷歌翻译
在学习新知识时,班级学习学习(CIL)与灾难性遗忘和无数据CIL(DFCIL)的斗争更具挑战性,而无需访问以前学过的课程的培训数据。尽管最近的DFCIL作品介绍了诸如模型反转以合成以前类的数据,但由于合成数据和真实数据之间的严重域间隙,它们无法克服遗忘。为了解决这个问题,本文提出了有关DFCIL的关系引导的代表学习(RRL),称为R-DFCIL。在RRL中,我们引入了关系知识蒸馏,以灵活地将新数据的结构关系从旧模型转移到当前模型。我们的RRL增强DFCIL可以指导当前的模型来学习与以前类的表示更好地兼容的新课程的表示,从而大大减少了在改善可塑性的同时遗忘。为了避免表示和分类器学习之间的相互干扰,我们在RRL期间采用本地分类损失而不是全球分类损失。在RRL之后,分类头将通过全球类平衡的分类损失进行完善,以解决数据不平衡问题,并学习新课程和以前类之间的决策界限。关于CIFAR100,Tiny-Imagenet200和Imagenet100的广泛实验表明,我们的R-DFCIL显着超过了以前的方法,并实现了DFCIL的新最新性能。代码可从https://github.com/jianzhangcs/r-dfcil获得。
translated by 谷歌翻译
深网络架构在不忘记以前的任务的情况下努力继续学习新任务。最近的趋势表明,基于参数扩展的动态架构可以在持续学习中有效地减少灾难性忘记。但是,现有方法通常需要在测试时需要任务标识符,需要复杂调整以平衡越来越多的参数,并且几乎不在任务中共享任何信息。结果,他们努力扩展到大量任务,而无需显着开销。在本文中,我们提出了一种基于专用编码器/解码器框架的变压器体系结构。批判性地,编码器和解码器在所有任务中共享。通过特殊令牌的动态扩展,我们专注于任务分发的解码器网络的各个向前。由于严格控制参数扩展,我们的策略缩小到大量任务,同时具有可忽略的内存和时间开销。此外,这种有效的策略不需要任何HyperParameter调整来控制网络的扩展。我们的模型在大型ImageNet100和ImageNet100上达到了Cifar100和最先进的表演,而参数比并发动态框架的参数越小。
translated by 谷歌翻译
Lifelong learning has attracted much attention, but existing works still struggle to fight catastrophic forgetting and accumulate knowledge over long stretches of incremental learning. In this work, we propose PODNet, a model inspired by representation learning. By carefully balancing the compromise between remembering the old classes and learning new ones, PODNet fights catastrophic forgetting, even over very long runs of small incremental tasks -a setting so far unexplored by current works. PODNet innovates on existing art with an efficient spatialbased distillation-loss applied throughout the model and a representation comprising multiple proxy vectors for each class. We validate those innovations thoroughly, comparing PODNet with three state-of-the-art models on three datasets: CIFAR100, ImageNet100, and ImageNet1000. Our results showcase a significant advantage of PODNet over existing art, with accuracy gains of 12.10, 6.51, and 2.85 percentage points, respectively. 5
translated by 谷歌翻译
视觉变压器(VIT)的几乎没有射击的学习能力很少进行,尽管有很大的需求。在这项工作中,我们从经验上发现,使用相同的少数学习框架,例如\〜元基线,用VIT模型代替了广泛使用的CNN特征提取器,通常严重损害了几乎没有弹药的分类性能。此外,我们的实证研究表明,在没有归纳偏见的情况下,VIT通常会在几乎没有射击的学习方面学习低资格的令牌依赖性,在这些方案下,只有几个标记的培训数据可获得,这在很大程度上会导致上述性能降级。为了减轻这个问题,我们首次提出了一个简单而有效的几杆培训框架,即自我推广的监督(Sun)。具体而言,除了对全球语义学习的常规监督外,太阳还进一步预处理了少量学习数据集的VIT,然后使用它来生成各个位置特定的监督,以指导每个补丁令牌。此特定于位置的监督告诉VIT哪个贴片令牌相似或不同,因此可以加速令牌依赖的依赖学习。此外,它将每个贴片令牌中的本地语义建模,以提高对象接地和识别能力,以帮助学习可概括的模式。为了提高特定于位置的监督的质量,我们进一步提出了两种技术:〜1)背景补丁过滤以滤掉背景补丁并将其分配为额外的背景类别; 2)空间一致的增强,以引入足够的多样性以增加数据,同时保持生成的本地监督的准确性。实验结果表明,使用VITS的太阳显着超过了其他VIT的少量学习框架,并且是第一个获得比CNN最先进的效果更高的性能。
translated by 谷歌翻译
Vision Transformer (ViT) extends the application range of transformers from language processing to computer vision tasks as being an alternative architecture against the existing convolutional neural networks (CNN). Since the transformer-based architecture has been innovative for computer vision modeling, the design convention towards an effective architecture has been less studied yet. From the successful design principles of CNN, we investigate the role of spatial dimension conversion and its effectiveness on transformer-based architecture. We particularly attend to the dimension reduction principle of CNNs; as the depth increases, a conventional CNN increases channel dimension and decreases spatial dimensions. We empirically show that such a spatial dimension reduction is beneficial to a transformer architecture as well, and propose a novel Pooling-based Vision Transformer (PiT) upon the original ViT model. We show that PiT achieves the improved model capability and generalization performance against ViT. Throughout the extensive experiments, we further show PiT outperforms the baseline on several tasks such as image classification, object detection, and robustness evaluation. Source codes and ImageNet models are available at https://github.com/naver-ai/pit.
translated by 谷歌翻译
Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. These highperforming vision transformers are pre-trained with hundreds of millions of images using a large infrastructure, thereby limiting their adoption.In this work, we produce competitive convolutionfree transformers trained on ImageNet only using a single computer in less than 3 days. Our reference vision transformer (86M parameters) achieves top-1 accuracy of 83.1% (single-crop) on ImageNet with no external data.We also introduce a teacher-student strategy specific to transformers. It relies on a distillation token ensuring that the student learns from the teacher through attention, typically from a convnet teacher. The learned transformers are competitive (85.2% top-1 acc.) with the state of the art on ImageNet, and similarly when transferred to other tasks. We will share our code and models.
translated by 谷歌翻译
当以连续的方式学习新任务时,深层神经网络倾向于忘记他们以前学到的任务,这种现象称为灾难性遗忘。班级增量学习方法旨在通过记忆以前学到的任务的一些示例,并从中蒸馏出知识来解决此问题。但是,现有的方法努力平衡跨课程的性能,因为它们通常将模型过于最新任务。在我们的工作中,我们建议通过引入一种实现级别平衡性能的逐步学习(TKIL)的新型方法来解决这些挑战。该方法保留了各个类别的表示形式,并平衡了每个类别的准确性,因此可以更好地达到总体准确性和差异。 TKIL方法基于神经切线核(NTK),该神经网络将神经网络作为无限宽度极限的内核函数的收敛行为。在tkil中,特征层之间的梯度被视为这些层的表示之间的距离,可以定义为切线切线损失(GTK损耗),因此将其与平均重量一起最小化。这允许TKIL自动识别任务并在推理过程中快速适应它。具有各种增量学习设置的CIFAR-100和Imagenet数据集的实验表明,这些策略允许TKIL优于现有的最新方法。
translated by 谷歌翻译
视觉世界中新对象的不断出现对现实世界部署中当前的深度学习方法构成了巨大的挑战。由于稀有性或成本,新任务学习的挑战通常会加剧新类别的数据。在这里,我们探讨了几乎没有类别学习的重要任务(FSCIL)及其极端数据稀缺条件。理想的FSCIL模型都需要在所有类别上表现良好,无论其显示顺序或数据的匮乏。开放式现实世界条件也需要健壮,并可以轻松地适应始终在现场出现的新任务。在本文中,我们首先重新评估当前的任务设置,并为FSCIL任务提出更全面和实用的设置。然后,受到FSCIL和现代面部识别系统目标的相似性的启发,我们提出了我们的方法 - 增强角损失渐进分类或爱丽丝。在爱丽丝(Alice)中,我们建议使用角度损失损失来获得良好的特征。由于所获得的功能不仅需要紧凑,而且还需要足够多样化以维持未来的增量类别的概括,我们进一步讨论了类增强,数据增强和数据平衡如何影响分类性能。在包括CIFAR100,Miniimagenet和Cub200在内的基准数据集上的实验证明了爱丽丝在最新的FSCIL方法上的性能提高。
translated by 谷歌翻译