持续创作，加速成长！这是我参与「掘金日新计划 · 6 月更文挑战」的第16天，点击查看活动详情
ShowMeAI日报系列全新升级！覆盖AI人工智能 工具&框架 | 项目&代码 | 博文&分享 | 数据&资源 | 研究&论文 等方向。点击查看 历史文章列表，在公众号内订阅话题 #ShowMeAI资讯日报，可接收每日最新推送。点击 专题合辑&电子月刊 快速浏览各专题全集。
1.工具&框架
工具：InferenceDB - 面向实时机器学习推理的数据库，基于Kafka
tags: [机器学习，推理，数据库，Kafka]
'InferenceDB - Stream inferences of real-time ML models in production to any data lake' by Aporia
GitHub: github.com/aporia-ai/i…
工具库：FortuneSheet - 嵌入式Javascript电子表格库，提供了如Excel和Google Sheets的丰富功能
tags: [电子表格，JavaScript，FortuneSheet]
'FortuneSheet - A drop-in javascript spreadsheet library that provides rich features like Excel and Google Sheets'
GitHub: github.com/ruilisi/for…
工具库：PranaDB - 可水平扩展分布式流数据库
tags: [分布式，数据库]
'PranaDB - a distributed streaming database, designed from the outset to be horizontally scalable’ by Cash App
GitHub: github.com/cashapp/pra…
工具库：milli - Rust写的高性能搜索引擎Meilisearch核心
tags: [搜索引擎，核心]
'milli - Search through millions of documents in milliseconds'
GitHub: github.com/meilisearch…
工具箱：B-Box - 对抗性黑箱攻击工具箱，用于评估Pytorch深度学习模型鲁棒性
tags: [对抗，黑箱攻击，神经网络鲁棒性]
'B-Box: an adversarial black-box attack toolbox to evaluate the robustness of Deep Learning models in Pytorch.' by SCLBD
GitHub: github.com/SCLBD/Black…
工具库：Slideflow - 用于组织学图像分析的深度学习流水线，支持Tensorflow/Keras和PyTorch
tags: [图像分析，pipeiline]
'Slideflow - Deep learning pipeline for histology image analysis, with both Tensorflow and PyTorch support.' by James Dolezal
GitHub: github.com/jamesdoleza…
2.项目&代码
项目：基于GPT3构建的国际象棋引擎
tags: [GPT3，国际象棋，引擎]
'GPT3Chess - Creating a chess engine using GPT-3' by ShamzGuy
GitHub: github.com/ShamzGuy/GP…
项目：推荐系统CTR预估实现
tags: [推荐系统，CTR预估，神经网络]
'CTR_Algorithm - 一些经典的CTR算法的复现;
包含LR, FM, FFM, AFM, DeepFM，xDeepFM, PNN, DCN, DCNv2, DIFM, AutoInt, FiBiNet,AFN,ONN,DIN, DIEN ... （pytorch, tf2.0）' by luo han
GitHub: github.com/Prayforhanl…
3.博文&分享
分享：工业界的推荐系统 - 结合小红书的业务场景和内部实践，讲解主流的工业界推荐系统技术
tags: [推荐系统，小红书，讲解]
GitHub: github.com/wangshusen/…
教程：Google为数据工程师、机器学习工程师、云工程师、云架构师、数据分析师等提供的免费学习路线
tags: [数据工程，机器学习，云架构师，数据分析，google，学习路径]
学习路线 | Google Cloud Skills Boost”
Link: www.cloudskillsboost.google/paths
4.数据&资源
资源列表：图对抗学习文献集
tags: [图对抗学习，文献集，资源列表]
'Graph Adversarial Learning Literature - A curated list of adversarial attacks and defenses papers on graph-structured data.' by SafeGraph
GitHub: github.com/safe-graph/…
资源列表：Open Source Data Annotation & Labeling Tools，开源数据标注和标记工具列表
tags: [数据标注，标注工具，资源列表]
'Open Source Data Annotation & Labeling Tools - Open Source Data Annotation & Labelling Tools' by ZenML
GitHub: github.com/zenml-io/aw…
5.研究&论文
公众号回复关键字 日报，免费获取整理好的6月论文合辑。
论文：Prompting ELECTRA: Few-Shot Learning with Discriminative Pre-Trained Models
论文标题：Prompting ELECTRA: Few-Shot Learning with Discriminative Pre-Trained Models
论文时间：30 May 2022
所属领域：自然语言处理
对应任务：Few-Shot Learning，Text Infilling，少样本学习，文本填充
论文作者：Mengzhou Xia, Mikel Artetxe, Jingfei Du, Danqi Chen, Ves Stoyanov
论文简介：In this work, we adapt prompt-based few-shot learning to ELECTRA and show that it outperforms masked language models in a wide range of tasks./在这项工作中，我们将基于提示（prompt）的小样本学习应用于 ELECTRA，并表明它在广泛的任务中优于掩码语言模型。
论文摘要：Pre-trained masked language models successfully perform few-shot learning by formulating downstream tasks as text infilling. However, as a strong alternative in full-shot settings, discriminative pre-trained models like ELECTRA do not fit into the paradigm. In this work, we adapt prompt-based few-shot learning to ELECTRA and show that it outperforms masked language models in a wide range of tasks. ELECTRA is pre-trained to distinguish if a token is generated or original. We naturally extend that to prompt-based few-shot learning by training to score the originality of the target options without introducing new parameters. Our method can be easily adapted to tasks involving multi-token predictions without extra computation overhead. Analysis shows that ELECTRA learns distributions that align better with downstream tasks.
通过将下游任务制定为文本填充，预训练的掩码语言模型成功地执行了小样本学习。 然而，作为full-shot设置中的一个强有力的替代方案，像 ELECTRA 这样的判别性预训练模型并不适合该范式。 在我们本次工作中，我们将基于提示的小样本学习应用于 ELECTRA，并表明它在广泛的任务中优于掩码语言模型。 ELECTRA 经过预训练以区分token是生成的还是原始的。 我们通过训练自然地将其扩展到基于提示（prompt）的少样本学习，以在不引入新参数的情况下对目标选项的独创性进行评分。 我们的方法可以很容易地适应涉及多token预测的任务，而无需额外的计算开销。 分析表明，ELECTRA 学习的分布更符合下游任务。
论文：Infinite Recommendation Networks: A Data-Centric Approach
论文标题：Infinite Recommendation Networks: A Data-Centric Approach
论文时间：3 Jun 2022
所属领域：自然语言处理，推荐系统
对应任务：Information Retrieval，Recommendation Systems，信息检索，推荐系统
代码实现：github.com/noveens/inf… , github.com/noveens/dis…
论文作者：Noveen Sachdeva, Mehak Preet Dhaliwal, Carole-Jean Wu, Julian McAuley
论文简介：We leverage the Neural Tangent Kernel and its equivalence to training infinitely-wide neural networks to devise ∞-AE: an autoencoder with infinitely-wide bottleneck layers./我们利用神经切线内核及其对训练无限宽神经网络的等效性来设计 -AE：一种具有无限宽瓶颈层的自动编码器。
论文摘要：We leverage the Neural Tangent Kernel and its equivalence to training infinitely-wide neural networks to devise ∞-AE: an autoencoder with infinitely-wide bottleneck layers. The outcome is a highly expressive yet simplistic recommendation model with a single hyper-parameter and a closed-form solution. Leveraging ∞-AE's simplicity, we also develop Distill-CF for synthesizing tiny, high-fidelity data summaries which distill the most important knowledge from the extremely large and sparse user-item interaction matrix for efficient and accurate subsequent data-usage like model training, inference, architecture search, etc. This takes a data-centric approach to recommendation, where we aim to improve the quality of logged user-feedback data for subsequent modeling, independent of the learning algorithm. We particularly utilize the concept of differentiable Gumbel-sampling to handle the inherent data heterogeneity, sparsity, and semi-structuredness, while being scalable to datasets with hundreds of millions of user-item interactions. Both of our proposed approaches significantly outperform their respective state-of-the-art and when used together, we observe 96-105% of ∞-AE's performance on the full dataset with as little as 0.1% of the original dataset size, leading us to explore the counter-intuitive question: Is more data what you need for better recommendation?
我们利用神经切线内核及其对训练无限宽神经网络的等效性来设计∞-AE：一种具有无限宽瓶颈层的自动编码器。结果是一个高度表达但简单的推荐模型，具有单个超参数和一个封闭形式的解决方案。利用 ∞-AE 的简单性，我们还开发了 Distill-CF，用于合成微小的高保真数据摘要，从极大且稀疏的用户项目交互矩阵中提取最重要的知识，以实现高效和准确的后续数据使用，如模型训练，推理、架构搜索等。这采用以数据为中心的推荐方法，我们的目标是提高记录的用户反馈数据的质量，以便后续建模，独立于学习算法。我们特别利用可微 Gumbel 采样的概念来处理固有的数据异质性、稀疏性和半结构化，同时可扩展到具有数亿用户-项目交互的数据集。我们提出的两种方法都显着优于它们各自的最新技术，当一起使用时，我们观察到在只有原始数据集大小的 0.1% 的前提下可以取得96-105% 的全量数据∞-AE 性能 ，让我们开始思考违反直觉的问题：我们需要更多数据来获得更好的推荐吗？
论文：Torsional Diffusion for Molecular Conformer Generation
论文标题：Torsional Diffusion for Molecular Conformer Generation
论文时间：1 Jun 2022
所属领域：计算化学
对应任务：Molecular conformer generation，分子构象生成
论文作者：Bowen Jing, Gabriele Corso, Jeffrey Chang, Regina Barzilay, Tommi Jaakkola
论文简介：Molecular conformer generation is a fundamental task in computational chemistry./分子构象异构体生成是计算化学中的一项基本任务。
论文摘要：Molecular conformer generation is a fundamental task in computational chemistry. Several machine learning approaches have been developed, but none have outperformed state-of-the-art cheminformatics methods. We propose torsional diffusion, a novel diffusion framework that operates on the space of torsion angles via a diffusion process on the hypertorus and an extrinsic-to-intrinsic score model. On a standard benchmark of drug-like molecules, torsional diffusion generates superior conformer ensembles compared to machine learning and cheminformatics methods in terms of both RMSD and chemical properties, and is orders of magnitude faster than previous diffusion-based models. Moreover, our model provides exact likelihoods, which we employ to build the first generalizable Boltzmann generator. Code is available at github.com/gcorso/tors…
分子构象异构体生成是计算化学中的一项基本任务。 已经开发了几种机器学习方法，但没有一种方法优于最先进的化学信息学方法。 我们提出了扭转扩散，这是一种新颖的扩散框架，通过超环上的扩散过程和外到内评分模型在扭转角空间上运行。 在药物样分子的标准基准上，与机器学习和化学信息学方法相比，扭转扩散在 RMSD 和化学性质方面产生了更好的构象集合，并且比以前的基于扩散的模型快几个数量级。 此外，我们的模型提供了精确的可能性，我们用它来构建第一个可泛化的玻尔兹曼生成器。 代码可在 github.com/gcorso/tors… 获得。
论文：POODLE: Improving Few-shot Learning via Penalizing Out-of-Distribution Samples
论文标题：POODLE: Improving Few-shot Learning via Penalizing Out-of-Distribution Samples
论文时间：NeurIPS 2021
所属领域：计算机视觉
对应任务：Few-Shot Learning，少样本学习
论文作者：Duong Le, Khoi Nguyen, Quoc-Huy Tran, Rang Nguyen, Binh-Son Hua
论文简介：In this work, we propose to use out-of-distribution samples, i. e., unlabeled samples coming from outside the target classes, to improve few-shot learning./在这项工作中，我们建议使用分布外样本，即，来自目标类之外的未标记样本，以改进小样本学习。
论文摘要：In this work, we propose to use out-of-distribution samples, i.e., unlabeled samples coming from outside the target classes, to improve few-shot learning. Specifically, we exploit the easily available out-of-distribution samples to drive the classifier to avoid irrelevant features by maximizing the distance from prototypes to out-of-distribution samples while minimizing that of in-distribution samples (i.e., support, query data). Our approach is simple to implement, agnostic to feature extractors, lightweight without any additional cost for pre-training, and applicable to both inductive and transductive settings. Extensive experiments on various standard benchmarks demonstrate that the proposed method consistently improves the performance of pretrained networks with different architectures.
在这项工作中，我们建议使用分布外样本，即来自目标类之外的未标记样本，以改进少样本学习。 具体来说，我们利用易于获得的分布外样本来驱动分类器，通过最大化原型到分布外样本的距离同时最小化分布内样本（即支持、查询数据）的距离来避免不相关的特征 . 我们的方法易于实现，与特征提取器无关，重量轻，无需任何额外的预训练成本，并且适用于感应和感应设置。 在各种标准基准上进行的大量实验表明，所提出的方法稳定地提高了具有不同架构的预训练网络的性能。
论文：Thin-Plate Spline Motion Model for Image Animation
论文标题：Thin-Plate Spline Motion Model for Image Animation
论文时间：CVPR 2022
所属领域：计算机视觉
对应任务：Image Animation，Motion Estimation，Optical Flow Estimation，图像动画，运动估计，光流估计
论文作者：Jian Zhao, HUI ZHANG
论文简介：Firstly, we propose thin-plate spline motion estimation to produce a more flexible optical flow, which warps the feature maps of the source image to the feature domain of the driving image./首先，我们提出thin-plate样条运动估计来产生更灵活的光流，它将源图像的特征图扭曲到驱动图像的特征域。
论文摘要：Image animation brings life to the static object in the source image according to the driving video. Recent works attempt to perform motion transfer on arbitrary objects through unsupervised methods without using a priori knowledge. However, it remains a significant challenge for current unsupervised methods when there is a large pose gap between the objects in the source and driving images. In this paper, a new end-to-end unsupervised motion transfer framework is proposed to overcome such issue. Firstly, we propose thin-plate spline motion estimation to produce a more flexible optical flow, which warps the feature maps of the source image to the feature domain of the driving image. Secondly, in order to restore the missing regions more realistically, we leverage multi-resolution occlusion masks to achieve more effective feature fusion. Finally, additional auxiliary loss functions are designed to ensure that there is a clear division of labor in the network modules, encouraging the network to generate high-quality images. Our method can animate a variety of objects, including talking faces, human bodies, and pixel animations. Experiments demonstrate that our method performs better on most benchmarks than the state of the art with visible improvements in pose-related metrics.
图像动画根据驱动视频为源图像中的静态对象带来运动活力。最近的工作试图通过无监督方法在不使用先验知识的情况下对任意对象执行运动转移。然而，当源图像和驱动图像中的对象之间存在较大的位姿差距时，对于当前的无监督方法来说，这仍然是一个重大挑战。在本文中，提出了一种新的端到端无监督运动传递框架来解决这个问题。首先，我们提出thin-plate样条运动估计来产生更灵活的光流，它将源图像的特征图扭曲到驱动图像的特征域。其次，为了更真实地恢复缺失区域，我们利用多分辨率遮挡掩码来实现更有效的特征融合。最后，设计了额外的辅助损失函数，以确保网络模块中有明确的分工，鼓励网络生成高质量的图像。我们的方法可以为各种对象设置动画，包括说话的面孔、人体和像素动画。实验表明，我们的方法在大多数基准测试中的表现优于现有技术，并且在姿势相关指标方面有明显改进。
论文：DN-DETR: Accelerate DETR Training by Introducing Query DeNoising
论文标题：DN-DETR: Accelerate DETR Training by Introducing Query DeNoising
论文时间：CVPR 2022
所属领域：计算机视觉
对应任务：目标检测
代码实现：github.com/fengli-ust/… , github.com/IDEA-openso… , github.com/slongliu/da… , github.com/IDEA-openso…
论文作者：Feng Li, Hao Zhang, Shilong Liu, Jian Guo, Lionel M. Ni, Lei Zhang
论文简介：Our method is universal and can be easily plugged into any DETR-like methods by adding dozens of lines of code to achieve a remarkable improvement./我们的方法是通用的，可以通过添加数十行代码轻松插入任何类似 DETR 的方法中，以实现显着的改进。
论文摘要：We present in this paper a novel denoising training method to speedup DETR (DEtection TRansformer) training and offer a deepened understanding of the slow convergence issue of DETR-like methods. We show that the slow convergence results from the instability of bipartite graph matching which causes inconsistent optimization goals in early training stages. To address this issue, except for the Hungarian loss, our method additionally feeds ground-truth bounding boxes with noises into Transformer decoder and trains the model to reconstruct the original boxes, which effectively reduces the bipartite graph matching difficulty and leads to a faster convergence. Our method is universal and can be easily plugged into any DETR-like methods by adding dozens of lines of code to achieve a remarkable improvement. As a result, our DN-DETR results in a remarkable improvement (+1.9AP) under the same setting and achieves the best result (AP 43.4 and 48.6 with 12 and 50 epochs of training respectively) among DETR-like methods with ResNet-50 backbone. Compared with the baseline under the same setting, DN-DETR achieves comparable performance with 50% training epochs. Code is available at github.com/FengLi-ust/…
我们在本文中提出了一种新的去噪训练方法来加速 DETR（DEtection TRansformer）训练，并加深了对 DETR 类方法的缓慢收敛问题的理解。我们表明，缓慢收敛是由于二部图匹配的不稳定性导致早期训练阶段的优化目标不一致。为了解决这个问题，除了 Hungarian loss，我们的方法额外将带有噪声的 ground-truth 边界框输入到 Transformer 解码器中，并训练模型重建原始框，这有效地降低了二分图匹配难度并导致更快的收敛。我们的方法是通用的，可以通过添加数十行代码轻松插入任何类似 DETR 的方法中，以实现显着的改进。结果，我们的 DN-DETR 在相同设置下取得了显着的改进（+1.9AP），并在使用 ResNet-50 骨干网络的 DETR 类方法中取得了最好的结果（AP 43.4 和 48.6，分别训练 12 和 50 个 epoch）。与相同设置下的基线相比，DN-DETR 以 50% 的训练轮次实现了卓越可比的性能。代码可在 github.com/FengLi-ust/… 获得。
论文：Benchmarking Robustness of 3D Point Cloud Recognition Against Common Corruptions
论文标题：Benchmarking Robustness of 3D Point Cloud Recognition Against Common Corruptions
论文时间：28 Jan 2022
所属领域：计算机视觉
对应任务：3D Point Cloud Classification，3D Point Cloud Data Augmentation，3D点云分类，3D点云数据增强
代码实现：github.com/jiachens/Mo… , github.com/tiangexiang… , github.com/dogyoonlee/… , github.com/dogyoonlee/… , github.com/RF5/CurveNe…
论文作者：Jiachen Sun, Qingzhao Zhang, Bhavya Kailkhura, Zhiding Yu, Chaowei Xiao, Z. Morley Mao
论文简介：Deep neural networks on 3D point cloud data have been widely used in the real world, especially in safety-critical applications./基于 3D 点云数据的深度神经网络已在现实世界中得到广泛应用，尤其是在安全关键型应用中。
论文摘要：Deep neural networks on 3D point cloud data have been widely used in the real world, especially in safety-critical applications. However, their robustness against corruptions is less studied. In this paper, we present ModelNet40-C, the first comprehensive benchmark on 3D point cloud corruption robustness, consisting of 15 common and realistic corruptions. Our evaluation shows a significant gap between the performances on ModelNet40 and ModelNet40-C for state-of-the-art (SOTA) models. To reduce the gap, we propose a simple but effective method by combining PointCutMix-R and TENT after evaluating a wide range of augmentation and test-time adaptation strategies. We identify a number of critical insights for future studies on corruption robustness in point cloud recognition. For instance, we unveil that Transformer-based architectures with proper training recipes achieve the strongest robustness. We hope our in-depth analysis will motivate the development of robust training strategies or architecture designs in the 3D point cloud domain. Our codebase and dataset are included in github.com/jiachens/Mo…
基于 3D 点云数据的深度神经网络已在现实世界中得到广泛应用，尤其是在安全关键型应用中。然而，它们对干扰的鲁棒性研究较少。在本文中，我们介绍了 ModelNet40-C，这是第一个关于 3D 点云损坏鲁棒性的综合基准，由 15 个常见和现实的损坏组成。我们的评估显示，ModelNet40 和 ModelNet40-C 在最先进 (SOTA) 模型上的性能存在显著差距。为了缩小差距，我们在评估了广泛的增强和测试时间适应策略后，通过结合 PointCutMix-R 和 TENT 提出了一种简单但有效的方法。我们为未来研究点云识别中的干扰破坏鲁棒性确定了一些重要的思路与方向。例如，我们揭示了具有适当训练方法的基于 Transformer 的架构实现了最强的鲁棒性。我们希望我们的深入分析能够促进 3D 点云领域中稳健的训练策略或架构设计的发展。我们的代码库和数据集包含在 github.com/jiachens/Mo…
论文：SimSwap: An Efficient Framework For High Fidelity Face Swapping
论文标题：SimSwap: An Efficient Framework For High Fidelity Face Swapping
论文时间：11 Jun 2021
所属领域：计算机视觉
对应任务：AI换脸
代码实现：github.com/neuralchen/… , github.com/woctezuma/S…
论文作者：Renwang Chen, Xuanhong Chen, Bingbing Ni, Yanhao Ge
论文简介：In contrast to previous approaches that either lack the ability to generalize to arbitrary identity or fail to preserve attributes like facial expression and gaze direction, our framework is capable of transferring the identity of an arbitrary source face into an arbitrary target face while preserving the attributes of the target face./与之前的方法要么缺乏泛化到任意身份的能力，要么无法保留面部表情和注视方向等属性，我们的框架能够将任意源人脸的身份转换为任意目标人脸，同时保留目标人脸的属性。
论文摘要：We propose an efficient framework, called Simple Swap (SimSwap), aiming for generalized and high fidelity face swapping. In contrast to previous approaches that either lack the ability to generalize to arbitrary identity or fail to preserve attributes like facial expression and gaze direction, our framework is capable of transferring the identity of an arbitrary source face into an arbitrary target face while preserving the attributes of the target face. We overcome the above defects in the following two ways. First, we present the ID Injection Module (IIM) which transfers the identity information of the source face into the target face at feature level. By using this module, we extend the architecture of an identity-specific face swapping algorithm to a framework for arbitrary face swapping. Second, we propose the Weak Feature Matching Loss which efficiently helps our framework to preserve the facial attributes in an implicit way. Extensive experiments on wild faces demonstrate that our SimSwap is able to achieve competitive identity performance while preserving attributes better than previous state-of-the-art methods. The code is already available on github: github.com/neuralchen/…
我们提出了一个高效的框架，称为简单交换（SimSwap），旨在实现广义和高保真人脸换脸。与之前的方法要么缺乏泛化到任意身份的能力，要么无法保留面部表情和注视方向等属性，我们的框架能够将任意源人脸的身份转换为任意目标人脸，同时保留目标人脸的属性。我们通过以下两种方式克服了上述缺陷：首先，我们提出了 ID 注入模块 (IIM)，它在特征级别将源人脸的身份信息传输到目标人脸。通过使用这个模块，我们将特定身份的人脸换脸算法的架构扩展到任意人脸换脸的框架。其次，我们提出了弱特征匹配损失，它有效地帮助我们的框架以隐式方式保留面部属性。对大量面孔的广泛实验表明，我们的 SimSwap 能够实现具有竞争力的身份性能，同时比以前的最先进方法更好地保留面部属性。该代码已在 github 上提供：github.com/neuralchen/…
我们是 ShowMeAI，致力于传播AI优质内容，分享行业解决方案，用知识加速每一次技术成长！点击查看 历史文章列表，在公众号内订阅话题 #ShowMeAI资讯日报，可接收每日最新推送。点击 专题合辑&电子月刊 快速浏览各专题全集。