Light guide plates are essential optical components widely used in a diverse range of applications ranging from medical lighting fixtures to back-lit TV displays. In this work, we introduce a fully-integrated, high-throughput, high-performance deep learning-driven workflow for light guide plate surface visual quality inspection (VQI) tailored for real-world manufacturing environments. To enable automated VQI on the edge computing within the fully-integrated VQI system, a highly compact deep anti-aliased attention condenser neural network (which we name LightDefectNet) tailored specifically for light guide plate surface defect detection in resource-constrained scenarios was created via machine-driven design exploration with computational and "best-practices" constraints as well as L_1 paired classification discrepancy loss. Experiments show that LightDetectNet achieves a detection accuracy of ~98.2% on the LGPSDD benchmark while having just 770K parameters (~33X and ~6.9X lower than ResNet-50 and EfficientNet-B0, respectively) and ~93M FLOPs (~88X and ~8.4X lower than ResNet-50 and EfficientNet-B0, respectively) and ~8.8X faster inference speed than EfficientNet-B0 on an embedded ARM processor. As such, the proposed deep learning-driven workflow, integrated with the aforementioned LightDefectNet neural network, is highly suited for high-throughput, high-performance light plate surface VQI within real-world manufacturing environments.
translated by 谷歌翻译
随着越来越多的深度学习对在设备上的Tinyml应用程序的采用,人们对对边缘进行优化的更有效的神经网络骨架的需求不断增加。最近,注意力冷凝器网络的引入导致低英寸,高效,自我发挥的神经网络,在准确性和速度之间取得了强大的平衡。在这项研究中,我们介绍了一种新的更快的注意力冷凝器设计,称为双感应注意力冷凝器,以实现更多的冷凝特征嵌入。我们进一步采用了机器驱动的设计探索策略,该策略施加了最佳实践设计限制,以提高效率和稳健性,以产生骨干的宏观构造结构。与其他几个其他最先进的有效骨架相比,所得的主链(我们命名为“参加”)在嵌入式ARM处理器上的推理吞吐量明显更高(以较高的精度和速度比FB-NET C快> 10倍)小型型号尺寸(以较高的速度和类似的精度小于OFA-62小1.47倍),并且准确性(以更高速度的ImageNet上的MobileVit Xs高1.1%)。这些有希望的结果表明,探索不同的有效体系结构设计和自我注意力的机制可以为Tinyml应用带来有趣的新构建块。
translated by 谷歌翻译
现代深度神经网络必须展示最先进的准确性,同时表现出低延迟和能耗。因此,在生成新架构时,神经结构搜索(NAS)算法将这两个约束考虑在内。但是,诸如延迟的效率度量通常是依赖于需要NAS算法来测量或预测架构延迟的硬件。测量每个评估架构的延迟增加了NAS过程的大量时间。在这里,我们将微处理器提出了一个先验的延迟估计枫木,其不依赖于传输学习或域适应,而是通过在训练期间结合先前的硬件特征来推广到新硬件。枫木利用新的定量策略来通过测量相关的硬件性能度量来表征底层微处理器,产生细粒度和富有效应硬件描述符。此外,所提出的枫木从CPU和GPU之间的紧密耦合I / O以及它们在从CPU中测量GPU硬件的微处理器性能硬件计数器时预测GPU上的DNN延迟的依赖性。通过这种定量策略作为硬件描述符,Maple可以通过一些镜头适应策略概括到新硬件,其中少于3个样本,它具有超过最先进的方法的3%改进,需要多达10个样品。实验结果表明,随着最先进的方法,增加了几次喷射适应样品到10提高了精度12%。此外,据证明,与任何数量适应样品的相关基线相比,枫木呈现出8-10%的精度,平均相比。
translated by 谷歌翻译
制造过程中的一个关键方面是用于缺陷和缺陷的制造部件的视觉质量检测。只有人类的视觉检查可能非常耗时和费力,并且是一个重要的瓶颈,特别是对于高吞吐制造场景。鉴于深度学习领域的显着进展,自动化视觉质量检验可能导致制造过程中的高效和可靠地检测缺陷和缺陷。然而,深度学习驱动的视觉检查方法通常需要大量的计算资源,从而限制吞吐量,并充当瓶颈,以实现智能工厂的广泛采用。在这项研究中,我们调查了利用机器驱动的设计探索方法来创建TinyDefectNet,这是一种高度紧凑的深度卷积网络架构,适用于高通量制造视觉质量检验。 TinyDefectNet包括仅〜427k的参数,并且具有〜97米的计算复杂性,但实现了最先进的架构的检测准确性,用于在Neu缺陷基准数据集上进行表面缺陷检测的任务。因此,TinyDefectNet可以在52 $ \ times $较低的架构复杂度和11x较低的计算复杂度下实现相同的检测性能。此外,使用AMD Zendnn Accelerator库,在AMD EPYC 7R32上部署了TinyDefectNet在AMD EPY 7R32上部署了7.6倍的吞吐量更快的吞吐量。最后,进行了解释性的性能验证策略,以确保TinyDefectNet展出了正确的决策行为,以改善运营商和检查员对其使用的信任。
translated by 谷歌翻译
National Association of Securities Dealers Automated Quotations(NASDAQ) is an American stock exchange based. It is one of the most valuable stock economic indices in the world and is located in New York City \cite{pagano2008quality}. The volatility of the stock market and the influence of economic indicators such as crude oil, gold, and the dollar in the stock market, and NASDAQ shares are also affected and have a volatile and chaotic nature \cite{firouzjaee2022lstm}.In this article, we have examined the effect of oil, dollar, gold, and the volatility of the stock market in the economic market, and then we have also examined the effect of these indicators on NASDAQ stocks. Then we started to analyze the impact of the feedback on the past prices of NASDAQ stocks and its impact on the current price. Using PCA and Linear Regression algorithm, we have designed an optimal dynamic learning experience for modeling these stocks. The results obtained from the quantitative analysis are consistent with the results of the qualitative analysis of economic studies, and the modeling done with the optimal dynamic experience of machine learning justifies the current price of NASDAQ shares.
translated by 谷歌翻译
Recent advances in language modeling have enabled new conversational systems. In particular, it is often desirable for people to make choices among specified options when using such systems. We address the problem of reference resolution, when people use natural expressions to choose between real world entities. For example, given the choice `Should we make a Simnel cake or a Pandan cake?' a natural response from a non-expert may be indirect: `let's make the green one'. Reference resolution has been little studied with natural expressions, thus robustly understanding such language has large potential for improving naturalness in dialog, recommendation, and search systems. We create AltEntities (Alternative Entities), a new public dataset of entity pairs and utterances, and develop models for the disambiguation problem. Consisting of 42K indirect referring expressions across three domains, it enables for the first time the study of how large language models can be adapted to this task. We find they achieve 82%-87% accuracy in realistic settings, which while reasonable also invites further advances.
translated by 谷歌翻译
WSD (Word Sense Disambiguation) is the task of identifying which sense of a word is meant in a sentence or other segment of text. Researchers have worked on this task (e.g. Pustejovsky, 2002) for years but it's still a challenging one even for SOTA (state-of-the-art) LMs (language models). The new dataset, TempoWiC introduced by Loureiro et al. (2022b) focuses on the fact that words change over time. Their best baseline achieves 70.33% macro-F1. In this work, we use two different losses simultaneously to train RoBERTa-based classification models. We also improve our model by using another similar dataset to generalize better. Our best configuration beats their best baseline by 4.23% and reaches 74.56% macroF1.
translated by 谷歌翻译
Covid-19是一种攻击上呼吸道和肺部的新型病毒。它的人对人的传播性非常迅速,这在个人生活的各个方面都引起了严重的问题。尽管一些感染的人可能仍然完全无症状,但经常被目睹有轻度至重度症状。除此之外,全球成千上万的死亡案件表明,检测Covid-19是社区的紧急需求。实际上,这是在筛选医学图像(例如计算机断层扫描(CT)和X射线图像)的帮助下进行的。但是,繁琐的临床程序和大量的每日病例对医生构成了巨大挑战。基于深度学习的方法在广泛的医疗任务中表现出了巨大的潜力。结果,我们引入了一种基于变压器的方法,用于使用紧凑卷积变压器(CCT)自动从X射线图像中自动检测COVID-19。我们的广泛实验证明了该方法的疗效,精度为98%,比以前的作品表现优于先前的作品。
translated by 谷歌翻译
不平衡的数据(ID)是阻止机器学习(ML)模型以实现令人满意的结果的问题。 ID是一种情况,即属于一个类别的样本的数量超过另一个类别的情况,这使此类模型学习过程偏向多数类。近年来,为了解决这个问题,已经提出了几种解决方案,该解决方案选择合成为少数族裔类生成新数据,或者减少平衡数据的多数类的数量。因此,在本文中,我们研究了基于深神经网络(DNN)和卷积神经网络(CNN)的方法的有效性,并与各种众所周知的不平衡数据解决方案混合,这意味着过采样和降采样。为了评估我们的方法,我们使用了龙骨,乳腺癌和Z-Alizadeh Sani数据集。为了获得可靠的结果,我们通过随机洗牌的数据分布进行了100次实验。分类结果表明,混合的合成少数族裔过采样技术(SMOTE) - 正态化-CNN优于在24个不平衡数据集上达到99.08%精度的不同方法。因此,提出的混合模型可以应用于其他实际数据集上的不平衡算法分类问题。
translated by 谷歌翻译
翻译质量估计(QE)是预测机器翻译(MT)输出质量的任务,而无需任何参考。作为MT实际应用中的重要组成部分,这项任务已越来越受到关注。在本文中,我们首先提出了XLMRScore,这是一种基于使用XLM-Roberta(XLMR)模型计算的BertScore的简单无监督的QE方法,同时讨论了使用此方法发生的问题。接下来,我们建议两种减轻问题的方法:用未知令牌和预训练模型的跨语性对准替换未翻译的单词,以表示彼此之间的一致性单词。我们在WMT21 QE共享任务的四个低资源语言对上评估了所提出的方法,以及本文介绍的新的英语FARSI测试数据集。实验表明,我们的方法可以在两个零射击方案的监督基线中获得可比的结果,即皮尔森相关性的差异少于0.01,同时在所有低资源语言对中的平均低资源语言对中的无人看管竞争对手的平均水平超过8%的平均水平超过8%。 。
translated by 谷歌翻译