自从2010年以来，深度学习技术对语音，语言，视觉等子领域的推动，在语言和视觉跨模态交叉学科领域我们也取得了很多激动人心的进展，包括跨语言与图像的理解、推理和生成。多模态智能旨在融合多种模态的信息进行处理实现智能应用，在5G时代将会是重要的热点技术之一。最近IEEE Fellow何晓东和邓力等作者撰写关于多模态智能的综述论文《Multimodal Intelligence: Representation Learning, Information Fusion, and Applications》，详述了多模态智能研究进展，涵盖259篇参考文献，本文从学习多模态表示、多模态信号在不同层次上的融合以及多模态应用三个新角度对多模态深度学习的最新研究成果进行了综合分析。

2020原论文，simple Siamese

通过去除不相关和多余的特征，特征选择旨在找到具有良好泛化能力的原始特征的紧凑表示。 随着无标签数据的普及，无监督特征选择已显示出可有效减轻维数的诅咒，对于全面分析和理解无标签高维数据的无数至关重要，这是由于子空间聚类中低秩表示法的成功所致，我们提出了一种用于无监督特征选择的正则化自我表示（RSR）模型，其中每个特征都可以表示为其相关特征的线性组合。 通过使用L-2，L-1-范数来表征表示系数矩阵和表示残差矩阵，RSR有效地选择了代表性特征并确保了对异常值的鲁棒性。 如果某个特征很重要，则它将参与大多数其他特征的表示，从而导致出现大量的表示系数，反之亦然。 对合成数据和现实世界数据进行的实验分析表明，该方法可以有效地识别代表性特征，在聚类精度，冗余减少和分类精度方面优于许多最新的无监督特征选择方法。

机器学习的视频中文字幕 Model Representation

A combined feature representation of deep feature and hand-crafted features for person re-identification

Domain-Specific Modeling has been widely and successfully used in system design and modeling of specific areas. Due to informal definition of Domain-Specific Modeling Language (DSML) and Meta-Modeling Language (DSMML), the structural semantics of DSMLs and DSMMLs cannot be strictly described and the

Cross-Project Transfer Representation Learning for Vulnerable Function Discovery

大牛ELAD关于sparse representation的代码

This textbook introduces sparse and redundant representations with a focus on applications in signal and image processing. The theoretical and numerical foundations are tackled before the applications are discussed. Mathematical modeling for signal sources is discussed along with how to use the proper model for tasks such as denoising, restoration, separation, interpolation and extrapolation, compression, sampling, analysis and synthesis, detection, recognition, and more. The presentation is elegant and engaging. Sparse and Redundant Representations is intended for graduate students in applied mathematics and electrical engineering, as well as applied mathematicians, engineers, and researchers who are active in the fields of signal and image processing. * Introduces theoretical and numerical foundations before tackling applications * Discusses how to use the proper model for various situations * Introduces sparse and redundant representations * Focuses on applications in signal and image processing The field of sparse and redundant representation modeling has gone through a major revolution in the past two decades. This started with a series of algorithms for approximating the sparsest solutions of linear systems of equations, later to be followed by surprising theoretical results that guarantee these algorithms’ performance. With these contributions in place, major barriers in making this model practical and applicable were removed, and sparsity and redundancy became central, leading to state-of-the-art results in various disciplines. One of the main beneficiaries of this progress is the field of image processing, where this model has been shown to lead to unprecedented performance in various applications. This book provides a comprehensive view of the topic of sparse and redundant representation modeling, and its use in signal and image processing. It offers a systematic and ordered exposure to the theoretical foundations of this data model, the numerical aspec

Image Super-Resolution Via Sparse Representation的中文版本