日韩性视频-久久久蜜桃-www中文字幕-在线中文字幕av-亚洲欧美一区二区三区四区-撸久久-香蕉视频一区-久久无码精品丰满人妻-国产高潮av-激情福利社-日韩av网址大全-国产精品久久999-日本五十路在线-性欧美在线-久久99精品波多结衣一区-男女午夜免费视频-黑人极品ⅴideos精品欧美棵-人人妻人人澡人人爽精品欧美一区-日韩一区在线看-欧美a级在线免费观看

歡迎訪問 生活随笔!

生活随笔

當(dāng)前位置: 首頁 > 编程资源 > 编程问答 >内容正文

编程问答

NIPS 2016上22篇论文的实现汇集

發(fā)布時間:2023/12/13 编程问答 39 豆豆
生活随笔 收集整理的這篇文章主要介紹了 NIPS 2016上22篇论文的实现汇集 小編覺得挺不錯的,現(xiàn)在分享給大家,幫大家做個參考.

日前,LightOn CEO 兼聯(lián)合創(chuàng)始人 Igor Carron 在其博客上放出了其收集到的 NIPS 2016 論文的實現(xiàn)(一共 22 個)。他寫道:「在 Reddit 上,peterkuharvarduk 決定編譯所有來自 NIPS 2016 的可用實現(xiàn),我很高興他使用了『實現(xiàn)( implementation)』這個詞,因為這讓我可以快速搜索到這些項目。」除了 peterkuharvarduk 的推薦,這里的項目還包括 Reddit 其他用戶和 Carron 額外添加的一些新公布的實現(xiàn)。最終他還重點推薦了 GitXiv:http://www.gitxiv.com 。另外,在本文后面還附帶了機(jī)器之心關(guān)于 NIPS 2016 的文章列表,千萬不要錯過。

  1. 使用快速權(quán)重關(guān)注最近的過去(Using Fast Weights to Attend to the Recent Past)

  論文:https://arxiv.org/abs/1610.06258

  GitHub:https://github.com/ajarai/fast-weights

  2. 通過梯度下降來學(xué)習(xí)通過梯度下降的學(xué)習(xí)(Learning to learn by gradient descent by gradient descent)

  論文:https://arxiv.org/abs/1606.04474

  GitHub:https://github.com/deepmind/learning-to-learn

  3. R-FCN:通過基于區(qū)域的全卷積網(wǎng)絡(luò)的目標(biāo)檢測(R-FCN: Object Detection via Region-based Fully Convolutional Networks)

  論文:https://arxiv.org/abs/1605.06409

  GitHub:https://github.com/Orpine/py-R-FCN

  4. 用于 k-均值的快速和可證明的 Good Seedings(Fast and Provably Good Seedings for k-Means)

  論文:https://las.inf.ethz.ch/files/bachem16fast.pdf.

  GitHub:https://github.com/obachem/kmc2

  5. 如何訓(xùn)練生成對抗網(wǎng)絡(luò)(How to Train a GAN)

  GitHub:https://github.com/soumith/ganhacks

  6. Phased LSTM:為長的或基于事件的序列加速循環(huán)網(wǎng)絡(luò)訓(xùn)練(Phased LSTM: Accelerating Recurrent Network Training for Long or Event-based Sequences)

  論文:https://arxiv.org/abs/1610.09513

  GitHub:?https://github.com/dannyneil/public_plstm

  7. 生成對抗式模仿學(xué)習(xí)(Generative Adversarial Imitation Learning)

  論文:https://arxiv.org/abs/1606.03476

  GitHub:https://github.com/openai/imitation

  8. 對抗式多類分類:一個風(fēng)險最小化的角度(Adversarial Multiclass Classification: A Risk Minimization Perspective)

  論文:https://www.cs.uic.edu/~rfathony/pdf/fathony2016adversarial.pdf

  GitHub:https://github.com/rizalzaf/adversarial-multiclass

  9. 通過視頻預(yù)測的用于物理交互的無監(jiān)督學(xué)習(xí)(Unsupervised Learning for Physical Interaction through Video Prediction)

  論文:https://arxiv.org/abs/1605.07157

  GitHub:?https://github.com/tensorflow/models/tree/master/video_prediction

  10.權(quán)重規(guī)范化:一種加速深度神經(jīng)網(wǎng)絡(luò)訓(xùn)練的簡單重新參數(shù)化( Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks)

  論文:https://arxiv.org/abs/1602.07868

  GitHub:https://github.com/openai/weightnorm

  11. 全容量整體循環(huán)神經(jīng)網(wǎng)絡(luò)(Full-Capacity Unitary Recurrent Neural Networks)

  論文:https://arxiv.org/abs/1611.00035

  GitHub:https://github.com/stwisdom/urnn

  12. 帶有隨機(jī)層的序列神經(jīng)模型(Sequential Neural Models with Stochastic Layers)

  論文:https://arxiv.org/pdf/1605.07571.pdf

  GitHub:https://github.com/marcofraccaro/srnn

  13. 帶有快速局部化譜過濾的圖上的卷積神經(jīng)網(wǎng)絡(luò)(Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering)

  論文:https://arxiv.org/abs/1606.09375

  GitHub:https://github.com/mdeff/cnn_graph

  14. Interpretable Distribution Features with Maximum Testing Power

  論文:https://papers.nips.cc/paper/6148-interpretable-distribution-features-with-maximum-testing-power.pdf

  GitHub:https://github.com/wittawatj/interpretable-test/

  15. 使用神經(jīng)網(wǎng)絡(luò)組成圖模型,用于結(jié)構(gòu)化表征和快速推理(Composing graphical models with neural networks for structured representations and fast inference )

  論文:https://arxiv.org/abs/1603.06277

  GitHub:https://github.com/mattjj/svae

  16. 使用張量網(wǎng)絡(luò)的監(jiān)督學(xué)習(xí)(Supervised Learning with Tensor Networks)

  論文:https://arxiv.org/abs/1605.05775

  GitHub:https://github.com/emstoudenmire/TNML

  17. 使用貝葉斯條件密度估計的模擬模型的快速無ε推理(Fast ε-free Inference of Simulation Models with Bayesian Conditional Density Estimation)

  論文:https://arxiv.org/abs/1605.06376

  GitHub:https://github.com/gpapamak/epsilon_free_inference

  18. 用于概率程序的貝葉斯優(yōu)化(Bayesian Optimization for Probabilistic Programs)

  論文:http://www.robots.ox.ac.uk/~twgr/assets/pdf/rainforth2016BOPP.pdf

  GitHub:https://github.com/probprog/bopp

  19. PVANet:用于實施目標(biāo)檢測的輕權(quán)重深度神經(jīng)網(wǎng)絡(luò)(PVANet: Lightweight Deep Neural Networks for Real-time Object Detection)

  論文:https://arxiv.org/abs/1611.08588

  GitHub:https://github.com/sanghoon/pva-faster-rcnn

  20. 數(shù)據(jù)編程:快速創(chuàng)建大訓(xùn)練集(Data Programming: Creating Large Training Sets Quickly)

  論文:https://arxiv.org/abs/1605.07723

  代碼:?snorkel.stanford.edu

  21. 用于架構(gòu)學(xué)習(xí)的卷積神經(jīng)結(jié)構(gòu)(Convolutional Neural Fabrics for Architecture Learning)

  論文:https://arxiv.org/pdf/1606.02492.pdf

  GitHub:https://github.com/shreyassaxena/convolutional-neural-fabrics

  22. 價值迭代網(wǎng)絡(luò)(Value Iteration Networks)

  論文:https://arxiv.org/abs/1602.02867

  TensorFlow 實現(xiàn):https://github.com/TheAbhiKumar/tensorflow-value-iteration-networks

? ? ??原作者的 Theano 實現(xiàn):https://github.com/avivt/VIN


總結(jié)

以上是生活随笔為你收集整理的NIPS 2016上22篇论文的实现汇集的全部內(nèi)容,希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯,歡迎將生活随笔推薦給好友。