专栏首页AI研习社126篇殿堂级深度学习论文分类整理 从入门到应用(上)

126篇殿堂级深度学习论文分类整理 从入门到应用(上)

█ 如果你有非常大的决心从事深度学习,又不想在这一行打酱油,那么研读大牛论文将是不可避免的一步。而作为新人,你的第一个问题或许是:“论文那么多,从哪一篇读起?”

本文将试图解决这个问题——文章标题本来是:“从入门到绝望,无止境的深度学习论文”。请诸位备好道具,开启头悬梁锥刺股的学霸姿势。

开个玩笑。

但对非科班出身的开发者而言,读论文的确可以成为一件很痛苦的事。但好消息来了——为避免初学者陷入迷途苦海,昵称为 songrotek 的学霸在 GitHub 发布了他整理的深度学习路线图,分门别类梳理了新入门者最需要学习的 DL 论文,又按重要程度给每篇论文打上星星。

截至目前,这份 DL 论文路线图已在 GitHub 收获了近万颗星星好评,人气极高。雷锋网感到非常有必要对大家进行介绍。

闲话少说,该路线图根据以下四项原则而组织:

  • 从大纲到细节
  • 从经典到前沿
  • 从一般到具体领域
  • 关注最新研究突破

作者注:有许多论文很新但非常值得一读。

1 深度学习历史和基础

1.0 书籍

█[0] Bengio, Yoshua, Ian J. Goodfellow, and Aaron Courville. "Deep learning." An MIT Press book. (2015). [pdf] (Ian Goodfellow 等大牛所著的教科书,乃深度学习圣经。你可以同时研习这本书以及以下论文) ★★★★★

地址:https://github.com/HFTrader/DeepLearningBook/raw/master/DeepLearningBook.pdf

1.1 调查

█[1] LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton. "Deep learning." Nature 521.7553 (2015): 436-444. [pdf] (三巨头做的调查) ★★★★★

地址:http://www.cs.toronto.edu/~hinton/absps/NatureDeepReview.pdf

1.2 深度置信网络 (DBN,深度学习前夜的里程碑)

█[2] Hinton, Geoffrey E., Simon Osindero, and Yee-Whye Teh. "A fast learning algorithm for deep belief nets." Neural computation 18.7 (2006): 1527-1554. [pdf] (深度学习前夜) ★★★

地址:http://www.cs.toronto.edu/~hinton/absps/ncfast.pdf

█[3] Hinton, Geoffrey E., and Ruslan R. Salakhutdinov. "Reducing the dimensionality of data with neural networks." Science 313.5786 (2006): 504-507. [pdf] (里程碑,展示了深度学习的前景) ★★★

地址:http://www.cs.toronto.edu/~hinton/science.pdf

1.3 ImageNet 的进化(深度学习从此萌发)

█[4] Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012. [pdf] (AlexNet, 深度学习突破) ★★★★★

地址:http://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf

█[5] Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." arXiv preprint arXiv:1409.1556 (2014). [pdf] (VGGNet,神经网络变得很深层) ★★★

地址:https://arxiv.org/pdf/1409.1556.pdf

█[6] Szegedy, Christian, et al. "Going deeper with convolutions." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2015. [pdf] (GoogLeNet) ★★★

地址:http://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Szegedy_Going_Deeper_With_2015_CVPR_paper.pdf

█[7] He, Kaiming, et al. "Deep residual learning for image recognition." arXiv preprint arXiv:1512.03385 (2015). [pdf](ResNet,特别深的神经网络, CVPR 最佳论文) ★★★★★

地址:https://arxiv.org/pdf/1512.03385.pdf

1.4 语音识别的进化

█[8] Hinton, Geoffrey, et al. "Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups." IEEE Signal Processing Magazine 29.6 (2012): 82-97. [pdf] (语音识别的突破) ★★★★

地址:http://cs224d.stanford.edu/papers/maas_paper.pdf

█[9] Graves, Alex, Abdel-rahman Mohamed, and Geoffrey Hinton. "Speech recognition with deep recurrent neural networks." 2013 IEEE international conference on acoustics, speech and signal processing. IEEE, 2013. [pdf] (RNN) ★★★

地址:http://arxiv.org/pdf/1303.5778.pdf

█[10] Graves, Alex, and Navdeep Jaitly. "Towards End-To-End Speech Recognition with Recurrent Neural Networks." ICML. Vol. 14. 2014. [pdf] ★★★

地址:http://www.jmlr.org/proceedings/papers/v32/graves14.pdf

█[11] Sak, Haşim, et al. "Fast and accurate recurrent neural network acoustic models for speech recognition." arXiv preprint arXiv:1507.06947 (2015). [pdf] (谷歌语音识别系统) ★★★

地址:http://arxiv.org/pdf/1507.06947

█[12] Amodei, Dario, et al. "Deep speech 2: End-to-end speech recognition in english and mandarin." arXiv preprint arXiv:1512.02595 (2015). [pdf] (百度语音识别系统) ★★★★

地址:https://arxiv.org/pdf/1512.02595.pdf

█[13] W. Xiong, J. Droppo, X. Huang, F. Seide, M. Seltzer, A. Stolcke, D. Yu, G. Zweig "Achieving Human Parity in Conversational Speech Recognition." arXiv preprint arXiv:1610.05256 (2016). [pdf] (最前沿的语音识别, 微软) ★★★★

地址:https://arxiv.org/pdf/1610.05256v1

研读以上论文之后,你将对深度学习历史、模型的基本架构(包括 CNN, RNN, LSTM)有一个基础的了解,并理解深度学习如何应用于图像和语音识别问题。接下来的论文,将带你深入探索深度学习方法、在不同领域的应用和前沿尖端技术。我建议,你可以根据兴趣和工作/研究方向进行选择性的阅读。

2 深度学习方法

2.1 模型

█[14] Hinton, Geoffrey E., et al. "Improving neural networks by preventing co-adaptation of feature detectors." arXiv preprint arXiv:1207.0580 (2012). [pdf] (Dropout) ★★★

地址:https://arxiv.org/pdf/1207.0580.pdf

█[15] Srivastava, Nitish, et al. "Dropout: a simple way to prevent neural networks from overfitting." Journal of Machine Learning Research 15.1 (2014): 1929-1958. [pdf] ★★★

地址:http://www.jmlr.org/papers/volume15/srivastava14a.old/source/srivastava14a.pdf

█[16] Ioffe, Sergey, and Christian Szegedy. "Batch normalization: Accelerating deep network training by reducing internal covariate shift." arXiv preprint arXiv:1502.03167 (2015). [pdf] (2015 年的杰出研究) ★★★★

地址:http://arxiv.org/pdf/1502.03167

█[17] Ba, Jimmy Lei, Jamie Ryan Kiros, and Geoffrey E. Hinton. "Layer normalization." arXiv preprint arXiv:1607.06450 (2016). [pdf] (Batch Normalization 的更新) ★★★★

地址:https://arxiv.org/pdf/1607.06450.pdf?utm_source=sciontist.com&utm_medium=refer&utm_campaign=promote

█[18] Courbariaux, Matthieu, et al. "Binarized Neural Networks: Training Neural Networks with Weights and Activations Constrained to+ 1 or−1." [pdf] (新模型,快) ★★★

地址:https://pdfs.semanticscholar.org/f832/b16cb367802609d91d400085eb87d630212a.pdf

█[19] Jaderberg, Max, et al. "Decoupled neural interfaces using synthetic gradients." arXiv preprint arXiv:1608.05343 (2016). [pdf] (训练方法的创新,研究相当不错) ★★★★★

地址:https://arxiv.org/pdf/1608.05343

█[20] Chen, Tianqi, Ian Goodfellow, and Jonathon Shlens. "Net2net: Accelerating learning via knowledge transfer." arXiv preprint arXiv:1511.05641 (2015). [pdf] (改进此前的训练网络,来缩短训练周期) ★★★

地址:https://arxiv.org/abs/1511.05641

█[21] Wei, Tao, et al. "Network Morphism." arXiv preprint arXiv:1603.01670 (2016). [pdf] (改进此前的训练网络,来缩短训练周期) ★★★

地址:https://arxiv.org/abs/1603.01670

2.2 优化 Optimization

█[22] Sutskever, Ilya, et al. "On the importance of initialization and momentum in deep learning." ICML (3) 28 (2013): 1139-1147. [pdf] (Momentum optimizer) ★★

地址:http://www.jmlr.org/proceedings/papers/v28/sutskever13.pdf

█[23] Kingma, Diederik, and Jimmy Ba. "Adam: A method for stochastic optimization." arXiv preprint arXiv:1412.6980 (2014). [pdf] (Maybe used most often currently) ★★★

地址:http://arxiv.org/pdf/1412.6980

█[24] Andrychowicz, Marcin, et al. "Learning to learn by gradient descent by gradient descent." arXiv preprint arXiv:1606.04474 (2016). [pdf] (Neural Optimizer,Amazing Work) ★★★★★

地址:https://arxiv.org/pdf/1606.04474

█[25] Han, Song, Huizi Mao, and William J. Dally. "Deep compression: Compressing deep neural network with pruning, trained quantization and huffman coding." CoRR, abs/1510.00149 2 (2015). [pdf] (ICLR best paper, new direction to make NN running fast,DeePhi Tech Startup) ★★★★★

地址:https://pdfs.semanticscholar.org/5b6c/9dda1d88095fa4aac1507348e498a1f2e863.pdf

█[26] Iandola, Forrest N., et al. "SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 1MB model size." arXiv preprint arXiv:1602.07360 (2016). [pdf] (Also a new direction to optimize NN,DeePhi Tech Startup) ★★★★

地址:http://arxiv.org/pdf/1602.07360

2.3 无监督学习/深度生成模型

█[27] Le, Quoc V. "Building high-level features using large scale unsupervised learning." 2013 IEEE international conference on acoustics, speech and signal processing. IEEE, 2013. [pdf] (里程碑, 吴恩达, 谷歌大脑, Cat) ★★★★

地址:http://arxiv.org/pdf/1112.6209.pdf&embed

█[28] Kingma, Diederik P., and Max Welling. "Auto-encoding variational bayes." arXiv preprint arXiv:1312.6114 (2013). [pdf](VAE) ★★★★

地址:http://arxiv.org/pdf/1312.6114

█[29] Goodfellow, Ian, et al. "Generative adversarial nets." Advances in Neural Information Processing Systems. 2014. [pdf](GAN,很酷的想法) ★★★★★

地址:http://papers.nips.cc/paper/5423-generative-adversarial-nets.pdf

█[30] Radford, Alec, Luke Metz, and Soumith Chintala. "Unsupervised representation learning with deep convolutional generative adversarial networks." arXiv preprint arXiv:1511.06434 (2015). [pdf] (DCGAN) ★★★★

地址:http://arxiv.org/pdf/1511.06434

█[31] Gregor, Karol, et al. "DRAW: A recurrent neural network for image generation." arXiv preprint arXiv:1502.04623 (2015). [pdf] (VAE with attention, 很出色的研究) ★★★★★

地址:http://jmlr.org/proceedings/papers/v37/gregor15.pdf

█[32] Oord, Aaron van den, Nal Kalchbrenner, and Koray Kavukcuoglu. "Pixel recurrent neural networks." arXiv preprint arXiv:1601.06759 (2016). [pdf] (PixelRNN) ★★★★

地址:http://arxiv.org/pdf/1601.06759

█[33] Oord, Aaron van den, et al. "Conditional image generation with PixelCNN decoders." arXiv preprint arXiv:1606.05328 (2016). [pdf] (PixelCNN) ★★★★

地址:https://arxiv.org/pdf/1606.05328

2.4 递归神经网络(RNN) / Sequence-to-Sequence Model

█[34] Graves, Alex. "Generating sequences with recurrent neural networks." arXiv preprint arXiv:1308.0850 (2013). [pdf] (LSTM, 效果很好,展示了 RNN 的性能) ★★★★

地址:http://arxiv.org/pdf/1308.0850

█[35] Cho, Kyunghyun, et al. "Learning phrase representations using RNN encoder-decoder for statistical machine translation." arXiv preprint arXiv:1406.1078 (2014). [pdf] (第一篇 Sequence-to-Sequence 的论文) ★★★★

地址:http://arxiv.org/pdf/1406.1078

█[36] Sutskever, Ilya, Oriol Vinyals, and Quoc V. Le. "Sequence to sequence learning with neural networks." Advances in neural information processing systems. 2014. [pdf] (杰出研究) ★★★★★

地址:http://papers.nips.cc/paper/5346-information-based-learning-by-agents-in-unbounded-state-spaces.pdf

█[37] Bahdanau, Dzmitry, KyungHyun Cho, and Yoshua Bengio. "Neural Machine Translation by Jointly Learning to Align and Translate." arXiv preprint arXiv:1409.0473 (2014). [pdf] ★★★★

地址:https://arxiv.org/pdf/1409.0473v7.pdf

█[38] Vinyals, Oriol, and Quoc Le. "A neural conversational model." arXiv preprint arXiv:1506.05869 (2015). [pdf] (Seq-to-Seq 聊天机器人) ★★★

地址:http://arxiv.org/pdf/1506.05869.pdf%20(http://arxiv.org/pdf/1506.05869.pdf)

2.5 神经网络图灵机

█[39] Graves, Alex, Greg Wayne, and Ivo Danihelka. "Neural turing machines." arXiv preprint arXiv:1410.5401 (2014). [pdf] (未来计算机的基础原型机) ★★★★★

地址:http://arxiv.org/pdf/1410.5401.pdf

█[40] Zaremba, Wojciech, and Ilya Sutskever. "Reinforcement learning neural Turing machines." arXiv preprint arXiv:1505.00521 362 (2015). [pdf] ★★★

地址:https://pdfs.semanticscholar.org/f10e/071292d593fef939e6ef4a59baf0bb3a6c2b.pdf

█[41] Weston, Jason, Sumit Chopra, and Antoine Bordes. "Memory networks." arXiv preprint arXiv:1410.3916 (2014). [pdf] ★★★

地址:http://arxiv.org/pdf/1410.3916

█[42] Sukhbaatar, Sainbayar, Jason Weston, and Rob Fergus. "End-to-end memory networks." Advances in neural information processing systems. 2015. [pdf] ★★★★

地址:http://papers.nips.cc/paper/5846-end-to-end-memory-networks.pdf

█[43] Vinyals, Oriol, Meire Fortunato, and Navdeep Jaitly. "Pointer networks." Advances in Neural Information Processing Systems. 2015. [pdf] ★★★★

地址:http://papers.nips.cc/paper/5866-pointer-networks.pdf

█[44] Graves, Alex, et al. "Hybrid computing using a neural network with dynamic external memory." Nature (2016). [pdf] (里程碑,把以上论文的想法整合了起来) ★★★★★

地址:https://www.dropbox.com/s/0a40xi702grx3dq/2016-graves.pdf

2.6 深度强化学习

█[45] Mnih, Volodymyr, et al. "Playing atari with deep reinforcement learning." arXiv preprint arXiv:1312.5602 (2013). [pdf]) (第一个以深度强化学习为题的论文) ★★★★

地址:http://arxiv.org/pdf/1312.5602.pdf

█[46] Mnih, Volodymyr, et al. "Human-level control through deep reinforcement learning." Nature 518.7540 (2015): 529-533. [pdf] (里程碑) ★★★★★

地址:https://storage.googleapis.com/deepmind-data/assets/papers/DeepMindNature14236Paper.pdf

█[47] Wang, Ziyu, Nando de Freitas, and Marc Lanctot. "Dueling network architectures for deep reinforcement learning." arXiv preprint arXiv:1511.06581 (2015). [pdf] (ICLR 最佳论文,很棒的想法) ★★★★

地址:http://arxiv.org/pdf/1511.06581

█[48] Mnih, Volodymyr, et al. "Asynchronous methods for deep reinforcement learning." arXiv preprint arXiv:1602.01783 (2016). [pdf] (前沿方法) ★★★★★

地址:http://arxiv.org/pdf/1602.01783

█[49] Lillicrap, Timothy P., et al. "Continuous control with deep reinforcement learning." arXiv preprint arXiv:1509.02971 (2015). [pdf] (DDPG) ★★★★

地址:http://arxiv.org/pdf/1509.02971

█[50] Gu, Shixiang, et al. "Continuous Deep Q-Learning with Model-based Acceleration." arXiv preprint arXiv:1603.00748 (2016). [pdf] (NAF) ★★★★

地址:http://arxiv.org/pdf/1603.00748

█[51] Schulman, John, et al. "Trust region policy optimization." CoRR, abs/1502.05477 (2015). [pdf] (TRPO) ★★★★

地址:http://www.jmlr.org/proceedings/papers/v37/schulman15.pdf

█[52] Silver, David, et al. "Mastering the game of Go with deep neural networks and tree search." Nature 529.7587 (2016): 484-489. [pdf] (AlphaGo) ★★★★★

地址:http://willamette.edu/~levenick/cs448/goNature.pdf

2.7 深度迁移学习 /终生学习 / 强化学习

█[53] Bengio, Yoshua. "Deep Learning of Representations for Unsupervised and Transfer Learning." ICML Unsupervised and Transfer Learning 27 (2012): 17-36. [pdf] (这是一个教程) ★★★

地址:http://www.jmlr.org/proceedings/papers/v27/bengio12a/bengio12a.pdf

█[54] Silver, Daniel L., Qiang Yang, and Lianghao Li. "Lifelong Machine Learning Systems: Beyond Learning Algorithms." AAAI Spring Symposium: Lifelong Machine Learning. 2013. [pdf] (对终生学习的简单讨论) ★★★

地址:http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.696.7800&rep=rep1&type=pdf

█[55] Hinton, Geoffrey, Oriol Vinyals, and Jeff Dean. "Distilling the knowledge in a neural network." arXiv preprint arXiv:1503.02531 (2015). [pdf] (大神们的研究) ★★★★

地址:http://arxiv.org/pdf/1503.02531

█[56] Rusu, Andrei A., et al. "Policy distillation." arXiv preprint arXiv:1511.06295 (2015). [pdf] (RL 领域) ★★★

地址:http://arxiv.org/pdf/1511.06295

█[57] Parisotto, Emilio, Jimmy Lei Ba, and Ruslan Salakhu★★★tdinov. "Actor-mimic: Deep multitask and transfer reinforcement learning." arXiv preprint arXiv:1511.06342 (2015). [pdf] (RL 领域) ★★★

地址:http://arxiv.org/pdf/1511.06342

█[58] Rusu, Andrei A., et al. "Progressive neural networks." arXiv preprint arXiv:1606.04671 (2016). [pdf] (杰出研究, 很新奇的想法) ★★★★★

地址:https://arxiv.org/pdf/1606.04671

2.8 One Shot 深度学习

█[59] Lake, Brenden M., Ruslan Salakhutdinov, and Joshua B. Tenenbaum. "Human-level concept learning through probabilistic program induction." Science 350.6266 (2015): 1332-1338. [pdf] (不含深度学习但值得一读) ★★★★★

地址:http://clm.utexas.edu/compjclub/wp-content/uploads/2016/02/lake2015.pdf

█[60] Koch, Gregory, Richard Zemel, and Ruslan Salakhutdinov. "Siamese Neural Networks for One-shot Image Recognition."(2015) [pdf] ★★★

地址:http://www.cs.utoronto.ca/~gkoch/files/msc-thesis.pdf

█[61] Santoro, Adam, et al. "One-shot Learning with Memory-Augmented Neural Networks." arXiv preprint arXiv:1605.06065 (2016). [pdf] (one shot 学习的基础一步) ★★★★

地址:http://arxiv.org/pdf/1605.06065

█[62] Vinyals, Oriol, et al. "Matching Networks for One Shot Learning." arXiv preprint arXiv:1606.04080 (2016). [pdf] ★★★

地址:https://arxiv.org/pdf/1606.04080

█[63] Hariharan, Bharath, and Ross Girshick. "Low-shot visual object recognition." arXiv preprint arXiv:1606.02819 (2016). [pdf] (通向更大规模数据的一步) ★★★★

地址:http://arxiv.org/pdf/1606.02819

本文分享自微信公众号 - AI研习社(okweiwu)

原文出处及转载信息见文内详细说明,如有侵权,请联系 yunjia_community@tencent.com 删除。

原始发表时间:2017-03-03

本文参与腾讯云自媒体分享计划,欢迎正在阅读的你也加入,一起分享。

我来说两句

0 条评论
登录 后参与评论

相关文章

  • DeepMind提出可微分逻辑编程,结合深度学习与符号程序优点

    夏乙 编译自 DeepMind Blog 量子位 出品 | 公众号 QbitAI 神经网络的强大功能有目共睹,但它往往需要大量与目标测试领域数据分布相似的训练数...

    量子位
  • Drive.ai首次展示技术实力:无人车夜雨中穿行山景城 | 视频

    美国加州的自动驾驶汽车创业公司Drive.ai,一直在努力使用深度学习技术,训练无人车系统。昨天,这家公司首度通过一段视频,披露了他们的工作进程。 在这段时长四...

    量子位
  • CNN实现“读脑术”,成功解码人脑视觉活动,准确率超50%

    【新智元导读】研究人员开发出以人脑为模型的深度学习算法,来破解人类大脑。相关研究发表在最新一期Cerebral Cortex,研究人员构建了一个大脑如何解码信息...

    新智元
  • 深度学习在推荐系统上的应用

    作者:陈仲铭 量子位 已获授权编辑发布 转载请联系原作者 深度学习最近大红大紫,深度学习的爆发使得人工智能进一步发展,阿里、腾讯、百度先后建立了自己的AI La...

    量子位
  • 对标谷歌TPU,比特大陆第一代深度学习专用处理器全球首发

    【新智元导读】人工智能时代,中国芯无论是技术还是市场都越来越精彩,继谷歌TPU之后,中国数字货币独角兽比特大陆,在今天的AI WORLD 2017世界人工智能大...

    新智元
  • 【Science】超越深度学习300倍, Vicarious发布生成视觉模型,LeCun批“这就是AI炒作的教科书”

    【新智元导读】最近大家都在探索“超越深度学习”的方法,“美国版DeepMind” Vicarious 近日在Science上发布的一项研究,使用不同于深度学习的...

    新智元
  • 【重磅】Hinton大神Capsule论文首次公布,深度学习基石CNN或被取代

    【新智元导读】Hinton要打造下一代CNN的Capsule细节终于通过一篇论文发布。本文带来详细介绍。此前,Hinton曾讨论了用“capsule”作为下一代...

    新智元
  • 22岁复旦学生拿下世界深度学习竞赛冠军:50层ResNet网络

    【新智元导读】 拥有世界上最大的开源车对车(V2V)网络的 Nexar 公布了第二届 Nexar 挑战赛的结果。来自复旦大学的Hengduo Li 拿下冠军。 ...

    新智元
  • 大数据与深度学习在一起:雅虎开源TensorFlowOnSpark

    量子位 李林 | 编译 雅虎宣布开源一个名为TensorFlowOnSpark的项目,支持对Apache Spark集群进行分布式TensorFlow训练和推断...

    量子位
  • 简化深度学习实践流程:新鲜出炉的TensorFlow项目模板来了

    林鳞 编译自 GitHub 量子位 出品 | 公众号 QbitAI 新的TensorFlow项目模板来了。 昨天,用户mrgemy95在Reddit上发帖,称这...

    量子位

扫码关注云+社区

领取腾讯云代金券