Deep Learning（深度學習）學習筆記（不斷更新）：
深度學習（Deep Learning）資料（不斷更新）：新增數據集，微信公眾號寫的更全些
為了您第一時間能獲取到最新資料，請關注微信公眾號：大數據技術宅
相關Paper（不斷更新）
筆者先從多個渠道整理了幾篇，后續邊看邊更新。
1、Densely Connected Convolutional Networks
2、Learning From Simulated and Unsupervised Images through Adversarial Training
3、Annotating Object Instance with a Polygon-RNN
4、YOLO9000: Better, Faster, Stronger
5、Computational Imaging on the Electric Grid
6、Object retrieval with large vocabularies and fast spatial matching
7、Improving Information Extraction by Acquiring External Evidence with Reinforcement Learning
8、Pointing the Unknown Words
9、LightRNN Memory and Computation-Efficient Recurrent Neural Network
10、Language Modeling with Gated Convolutional Networks
11、Recurrent neural network based language model
12、Extensions of Recurrent Neural Network Language Model
13、A guide to recurrent neural networks and backpropagation
14、Training Recurrent Neural Networks
15、Recurrent Neural Networks for Language Understanding
16、Empirical Evaluation and Combination of Advanced Language Modeling Techniques
17、Speech Recognition with Deep Recurrent Neural Networks
18、A fast learning algorithm for deep belief nets
19、Large Scale Distributed Deep Networks
20、Context Dependent Pretrained Deep Neural Networks fo Large Vocabulary Speech Recognition
21、An Empirical Study of Learning Rates in Deep Neural Networks for Speech Recognition
22、Deep Neural Networks for Acoustic Modeling in Speech Recognition
23、Deep Belief Networks Using Discriminative Features for Phone Recognition
24、Improving Deep Neural Networks For LVCSR using Rectified Linear Units and Dropout
25、Improved feature processing for Deep Neural Networks
26、Exploiting Sparseness in Deep Neural Networks fo Large Vocabulary Speech Recognition
27、Learning Features from Music Audio with Deep Belief Networks
28、Making Deep Belief Networks Effective for Large Vocabulary Continuous Speech Recognition
29、Robust Visual Recognition Using Multilayer Generative Neural Networks
30、Deep Convolutional Network Cascade for Facial Point Detection
31、ImageNet Classification with Deep Convolutional Neural Networks
32、Gradient-Based Learning Applied to Document Recognition
33、Stochastic Pooling for Regularization of Deep Convolutional Neural Networks
34、Best Practices for Convolutional Neural Networks Applied to Visual Document Analysis
35、Multi-GPU Training of ConvNets
36、Deep Learning For Signal And Information Processing
37、Deep Convex Net: A Scalable Architecture for Speech Pattern Classification
38、Improving Wideband Speech Recognition using Mixed-Bandwidth Training Data in CD-DNN-HMM
39、On Rectified Linear Units for Speech Processing
更新中。。。
相關書籍（不斷更新）
筆者剛着手學習，非大牛，不敢說“推薦”書籍，僅羅列所看的。
1、Deep Learning，出自Goodfellow、Bengio 和 Courville 三位大牛之手，筆者剛開始看，后續再對書籍作評論
如果需要《Deep Learning》中文電子版書籍，請后台回復“深度學習”獲取
更新中。。。
數據集（不斷更新）：
一、圖像數據集
1.MNIST:https://datahack.analyticsvidh ... gits/
MNIST是最受歡迎的深度學習數據集之一，這是一個手寫數字數據集，包含一組60,000個示例的訓練集和一個包含10,000個示例的測試集。這是一個很好的數據庫，用於在實際數據中嘗試學習技術和深度識別模式，同時可以在數據預處理中花費最少的時間和精力。
•大小： 50 MB
•記錄數量： 70,000張圖片被分成了10個組。
•SOTA： Capsules之間的動態路由
https://arxiv.org/pdf/1710.09829.pdf
2.MS-COCO:http://cocodataset.org/#home
COCO是一個大型的、豐富的物體檢測，分割和字幕數據集。它有幾個特點：
•對象分割；
•在上下文中可識別；
•超像素分割；
•330K圖像（> 200K標記）；
•150萬個對象實例；
•80個對象類別；
•91個類別；
•每張圖片5個字幕；
•有關鍵點的250,000人；
•大小：25 GB（壓縮）
•記錄數量： 330K圖像、80個對象類別、每幅圖像有5個標簽、25萬個關鍵點。
•SOTA：Mask R-CNN：https://arxiv.org/pdf/1703.06870.pdf
3.ImageNet：http://www.image-net.org/
ImageNet是根據WordNet層次結構組織的圖像數據集。WordNet包含大約100,000個單詞，ImageNet平均提供了大約1000個圖像來說明每個單詞。
大小：150GB
記錄數量：總圖像是大約是1,500,000，每個都有多個邊界框和相應的類標簽。
SOTA：深度神經網絡的聚合殘差變換。
https://arxiv.org/pdf/1611.05431.pdf
更多精彩內容，歡迎掃碼關注以下微信公眾號：大數據技術宅。大數據、AI從關注開始
