与其2D图像对应物相比，3D点云数据上的零射击学习是一个相关的未置换问题。 3D数据由于不可用的预训练特征提取模型而带来了ZSL的新挑战。为了解决这个问题，我们提出了一种及时引导的3D场景生成和监督方法，该方法可以增强3D数据以更好地学习网络，从而探索可见和看不见的对象的复杂相互作用。首先，我们以提示描述的某些方式合并了两个3D模型的点云。提示的行为就像描述每个3D场景的注释一样。后来，我们进行对比学习，以端到端的方式培训我们所提出的建筑。我们认为，与单​​个对象相比，3D场景可以更有效地关联对象，因为当对象出现在上下文中时，流行的语言模型（如Bert）可以实现高性能。我们提出的及时引导场景生成方法封装了数据扩展和基于及时的注释/字幕，以提高3D ZSL性能。我们已经在合成（ModelNet40，ModelNet10）和实扫描（ScanoJbectnn）3D对象数据集上实现了最新的ZSL和广义ZSL性能。

Automatic medical image classification is a very important field where the use of AI has the potential to have a real social impact. However, there are still many challenges that act as obstacles to making practically effective solutions. One of those is the fact that most of the medical imaging datasets have a class imbalance problem. This leads to the fact that existing AI techniques, particularly neural network-based deep-learning methodologies, often perform poorly in such scenarios. Thus this makes this area an interesting and active research focus for researchers. In this study, we propose a novel loss function to train neural network models to mitigate this critical issue in this important field. Through rigorous experiments on three independently collected datasets of three different medical imaging domains, we empirically show that our proposed loss function consistently performs well with an improvement between 2%-10% macro f1 when compared to the baseline models. We hope that our work will precipitate new research toward a more generalized approach to medical image classification.

We propose KnowGL, a tool that allows converting text into structured relational data represented as a set of ABox assertions compliant with the TBox of a given Knowledge Graph (KG), such as Wikidata. We address this problem as a sequence generation task by leveraging pre-trained sequence-to-sequence language models, e.g. BART. Given a sentence, we fine-tune such models to detect pairs of entity mentions and jointly generate a set of facts consisting of the full set of semantic annotations for a KG, such as entity labels, entity types, and their relationships. To showcase the capabilities of our tool, we build a web application consisting of a set of UI widgets that help users to navigate through the semantic data extracted from a given input text. We make the KnowGL model available at https://huggingface.co/ibm/knowgl-large.

正如GPT-3和T5所证明的那样，随着参数空间变得越来越大，变压器具有能力。但是，对于需要大量知识的任务，非参数存储器允许模型在计算成本和GPU内存需求的次线性增加中急剧增长。诸如RAG和Realm之类的最新模型已将检索引入条件生成。这些模型结合了从一系列语料库中的神经初始检索。我们基于这一研究，提出了RE2G，该研究将神经初始检索和重新融合到基于巴特的序列到序列的生成中。我们的阅读方法还允许从无与伦比分数的来源合并结果，从而实现BM25和神经初始检索的合奏。为了训练我们的系统端到端，我们引入了一种新颖的知识蒸馏变体，以在目标序列输出上仅使用地面真理来训练初始检索，重读者和生成。我们在四个不同的任务中发现了很大的收益：零击插槽填充，问答，事实检查和对话，相对增长了9％至34％，比以前的苏格兰短裙排行榜上的最先前的排行榜相比。我们将代码作为开源提供，网址为https://github.com/ibm/kgi-slot-filling/tree/re2g。

研究部门在组织中推动创新的重要作用。随着速度和量的信息增长，绘制见解，跟随趋势，保持新的研究以及制定策略的配制策略越来越越来越具有挑战性。在本文中，我们介绍了一个用例，即公司研究界如何利用语义网络技术来诱导从结构化和文本数据中诱导统一的知识图，通过整合与研究项目相关的社区使用的各种应用程序，学术论文，学术论文，数据集，成就和认可。为了使应用程序开发人员更容易访问知识图，我们确定了一组通用模式，用于利用诱导的知识并将其视为API。这些模式是从用户研究中诞生的，这些模式确定了最有价值的用例或用户疼痛点要缓解。我们概述了两个不同的方案：用于业务使用的建议和分析。我们将详细讨论这些方案，并针对实体建议提供经验评估。所使用的方法和从这项工作中学到的教训可以应用于面临类似挑战的其他组织。

很少有类别的课堂学习（FSCIL）旨在使用一些示例逐步微调模型（在基础课上培训），而不忘记先前的培训。最近的工作主要解决了2D图像。但是，由于相机技术的发展，3D点云数据比以往任何时候都更可用，这需要考虑3D数据的FSCIL。本文介绍了3D域中的FSCIL。除了灾难性忘记过去的知识和过度贴合数据的众所周知的问题外，3D FSCIL还可以带来更新的挑战。例如，基类可能在现实情况下包含许多合成实例。相比之下，新型类​​别只有少数几个实际扫描的样本（来自RGBD传感器）以增量步骤获得。由于数据从合成到真实的变化，FSCIL会承受其他挑战，以后的增量步骤降低了性能。我们尝试使用微莎普（正交基矢量）来解决此问题，并使用预定义的一组规则来描述任何3D对象。它支持逐步训练，几乎没有示例将合成与真实数据变化最小化。我们使用流行的合成数据集（ModelNet和Shapenet）和3D实范围的数据集（ScanoBjectNN和CO3D）为3D FSCIL提供新的测试协议。通过比较最先进的方法，我们确定了3D域中方法的有效性。

在本文中，我们介绍了一个系统，以展示最新的最新检索增强生成模型的功能，该模型接受了知识密集型语言任务的培训，例如插槽填充，开放式域问题答案，对话和事实检查。此外，鉴于用户查询，我们显示如何将这些不同模型的输出组合在一起以互相盘问彼此的输出。特别是，我们展示了使用问题答案模型如何提高对话的准确性。我们还将发布演示中使用的所有模型作为本文的贡献。一个简短的视频，展示了该系统，请访问https://ibm.box.com/v/emnlp2022-demo。

The usage of technologically advanced devices has seen a boom in many domains, including education, automation, and healthcare; with most of the services requiring Internet connectivity. To secure a network, device identification plays key role. In this paper, a device fingerprinting (DFP) model, which is able to distinguish between Internet of Things (IoT) and non-IoT devices, as well as uniquely identify individual devices, has been proposed. Four statistical features have been extracted from the consecutive five device-originated packets, to generate individual device fingerprints. The method has been evaluated using the Random Forest (RF) classifier and different datasets. Experimental results have shown that the proposed method achieves up to 99.8% accuracy in distinguishing between IoT and non-IoT devices and over 97.6% in classifying individual devices. These signify that the proposed method is useful in assisting operators in making their networks more secure and robust to security breaches and unauthorized access.

Objective: Despite numerous studies proposed for audio restoration in the literature, most of them focus on an isolated restoration problem such as denoising or dereverberation, ignoring other artifacts. Moreover, assuming a noisy or reverberant environment with limited number of fixed signal-to-distortion ratio (SDR) levels is a common practice. However, real-world audio is often corrupted by a blend of artifacts such as reverberation, sensor noise, and background audio mixture with varying types, severities, and duration. In this study, we propose a novel approach for blind restoration of real-world audio signals by Operational Generative Adversarial Networks (Op-GANs) with temporal and spectral objective metrics to enhance the quality of restored audio signal regardless of the type and severity of each artifact corrupting it. Methods: 1D Operational-GANs are used with generative neuron model optimized for blind restoration of any corrupted audio signal. Results: The proposed approach has been evaluated extensively over the benchmark TIMIT-RAR (speech) and GTZAN-RAR (non-speech) datasets corrupted with a random blend of artifacts each with a random severity to mimic real-world audio signals. Average SDR improvements of over 7.2 dB and 4.9 dB are achieved, respectively, which are substantial when compared with the baseline methods. Significance: This is a pioneer study in blind audio restoration with the unique capability of direct (time-domain) restoration of real-world audio whilst achieving an unprecedented level of performance for a wide SDR range and artifact types. Conclusion: 1D Op-GANs can achieve robust and computationally effective real-world audio restoration with significantly improved performance. The source codes and the generated real-world audio datasets are shared publicly with the research community in a dedicated GitHub repository1.

Multi-view projection techniques have shown themselves to be highly effective in achieving top-performing results in the recognition of 3D shapes. These methods involve learning how to combine information from multiple view-points. However, the camera view-points from which these views are obtained are often fixed for all shapes. To overcome the static nature of current multi-view techniques, we propose learning these view-points. Specifically, we introduce the Multi-View Transformation Network (MVTN), which uses differentiable rendering to determine optimal view-points for 3D shape recognition. As a result, MVTN can be trained end-to-end with any multi-view network for 3D shape classification. We integrate MVTN into a novel adaptive multi-view pipeline that is capable of rendering both 3D meshes and point clouds. Our approach demonstrates state-of-the-art performance in 3D classification and shape retrieval on several benchmarks (ModelNet40, ScanObjectNN, ShapeNet Core55). Further analysis indicates that our approach exhibits improved robustness to occlusion compared to other methods. We also investigate additional aspects of MVTN, such as 2D pretraining and its use for segmentation. To support further research in this area, we have released MVTorch, a PyTorch library for 3D understanding and generation using multi-view projections.