首页 > 其他分享 >Robust Deep Reinforcement Learning against Adversarial Perturbations on State Observations

Robust Deep Reinforcement Learning against Adversarial Perturbations on State Observations

时间:2023-05-05 11:11:35浏览次数:39  
标签:Adversarial against State Perturbations Reinforcement Deep

郑重声明:原文参见标题,如有侵权,请联系作者,将会撤销发布!

NeurIPS 2020

 

标签:Adversarial,against,State,Perturbations,Reinforcement,Deep
From: https://www.cnblogs.com/lucifer1997/p/17373551.html

相关文章

  • 论文解读《Interpolated Adversarial Training: Achieving robust neural networks wi
    论文信息论文标题:InterpolatedAdversarialTraining:Achievingrobustneuralnetworkswithoutsacrificingtoomuchaccuracy论文作者:AlexLambVikasVermaKenjiKawaguchiAlexanderMatyaskoSavyaKhoslaJuhoKannalaYoshuaBengio论文来源:2022NeuralNetworks论文地址:dow......
  • 论文阅读-sparse gpu kernels for deep learning
    论文地址:https://ieeexplore.ieee.org/document/9355309源码地址:https://github.com/google-research/sputnik背景深度神经网络由大量的矩阵乘法运算和卷积运算组成,这些运算中使用的矩阵可以转化成稀疏矩阵,同时不损失模型的精度。这样就可以在准确率不变的情况下提升浮点运算效......
  • DeepMind:用 GNN 学习通用推理算法
    文|智商掉了一地小孩子才做选择,我的模型全!都!要!近年来,基于深度神经网络的机器学习系统取得了巨大进步,尤其是在以感知为主的任务上。这一领域表现突出的模型通常要在分布中进行泛化,意味着它们的训练和验证集代表了测试输入的预期分布。相比之下,要真正掌握由推理主导的任务,即使是在......
  • 推翻OpenAI结论,DeepMind重新定义预训练的参数和规模关系!
    文|王思若前言从20年开始,“最大语言模型”的桂冠被各大研究机构和科技公司竞相追逐,堆砌参数,猛上算力,开启了“大炼丹”时代,模型参数量仿佛越大越好,甚至GPT-4模型参数量将超过100万亿的传闻甚嚣尘上。当把视角落在今年下半年,大模型的“军备竞赛”似乎戛然而止,22年4月,Google发布了5400......
  • Deep Dynamics Models for Learning Dexterous Manipulation
    发表时间:2019(CoRL2019)文章要点:文章提出了一个onlineplanningwithdeepdynamicsmodels(PDDM)的算法来学习Dexterousmulti-fingeredhands,大概意思就是学习拟人的灵活的手指操控技巧。大概思路就是结合uncertainty-awareneuralnetworkmodels和gradient-freetrajecto......
  • Cluster-GCN An Efficient Algorithm for Training Deep Convolution Networks
    目录概符号说明MotivationCluster-GCN代码ChiangW.,LiuX.,SiS.,LiY.,BengioS.andHsiehC.Cluster-GCN:Anefficientalgorithmfortrainingdeepandlargegraphconvolutionalnetworks.KDD,2019.概以往的GraphSage,FastGCN等方法,虽然能够实现mini-b......
  • Cycle GAN:Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial
    paper:https://arxiv.org/pdf/1703.10593.pdf[2017]code参考:https://github.com/junyanz/pytorch-CycleGAN-and-pix2pixhttps://zhuanlan.zhihu.com/p/79221194https://blog.csdn.net/fangjin_kl/article/details/128117396https://www.bilibili.com/video/BV1kb4y197P......
  • Deep-Learning-Based Spatio-Temporal-Spectral Integrated Fusion of Heterogeneous
    Deep-Learning-BasedSpatio-Temporal-SpectralIntegratedFusionofHeterogeneousRemoteSensingImagesabstract为了解决STF中的生成heterogeneousimages问题:为此,本文首次提出了一种基于新型深度残差循环生成对抗网络(GAN)的异构集成框架。所提出的网络由前向融合部......
  • Adversarial Robust Deep Reinforcement Learning Requires Redefining Robustness
    郑重声明:原文参见标题,如有侵权,请联系作者,将会撤销发布! ......
  • 论文解读(VAT)《Virtual Adversarial Training: A Regularization Method for Supervise
    论文信息论文标题:VirtualAdversarialTraining:ARegularizationMethodforSupervisedandSemi-SupervisedLearning论文作者:TakeruMiyato,S.Maeda,MasanoriKoyama,S.Ishii论文来源:2020ECCV论文地址:download 论文代码:download视屏讲解:click1前言提出问题:在......