RNN and LSTM saliency Predection Scene Label

http://handong1587.github.io/deep_learning/2015/10/09/rnn-and-lstm.html  //RNN and LSTM

http://handong1587.github.io/deep_learning/2015/10/09/saliency-prediction.html //saliency Predection

http://handong1587.github.io/deep_learning/2015/10/09/scene-labeling.html //Scene Label

 

RNN and LSTM

 Published: 09 Oct 2015  Category: deep_learning

Types of RNN

1) Plain Tanh Recurrent Nerual Networks

2) Gated Recurrent Neural Networks (GRU)

3) Long Short-Term Memory (LSTM)

Tutorials

A Beginner’s Guide to Recurrent Networks and LSTMs

http://deeplearning4j.org/lstm.html

A Deep Dive into Recurrent Neural Nets

http://nikhilbuduma.com/2015/01/11/a-deep-dive-into-recurrent-neural-networks/

Long Short-Term Memory: Tutorial on LSTM Recurrent Networks

http://people.idsia.ch/~juergen/lstm/index.htm

LSTM implementation explained

http://apaszke.github.io/lstm-explained.html

Recurrent Neural Networks Tutorial

• Part 1(Introduction to RNNs): http://www.wildml.com/2015/09/recurrent-neural-networks-tutorial-part-1-introduction-to-rnns/
• Part 2(Implementing a RNN using Python and Theano):http://www.wildml.com/2015/09/recurrent-neural-networks-tutorial-part-2-implementing-a-language-model-rnn-with-python-numpy-and-theano/
• Part 3(Understanding the Backpropagation Through Time (BPTT) algorithm):http://www.wildml.com/2015/10/recurrent-neural-networks-tutorial-part-3-backpropagation-through-time-and-vanishing-gradients/
• Part 4(Implementing a GRU/LSTM RNN): http://www.wildml.com/2015/10/recurrent-neural-network-tutorial-part-4-implementing-a-grulstm-rnn-with-python-and-theano/

Understanding LSTM Networks

• blog: http://colah.github.io/posts/2015-08-Understanding-LSTMs/
• ZH: http://www.jianshu.com/p/9dc9f41f0b29

Recurrent Neural Networks in DL4J

http://deeplearning4j.org/usingrnns.html

Train RNN

A Simple Way to Initialize Recurrent Networks of Rectified Linear Units

• arxiv: http://arxiv.org/abs/1504.00941
• gitxiv: http://gitxiv.com/posts/7j5JXvP3kn5Jf8Waj/irnn-experiment-with-pixel-by-pixel-sequential-mnist
• github: https://github.com/fchollet/keras/blob/master/examples/mnist_irnn.py
• github: https://gist.github.com/GabrielPereyra/353499f2e6e407883b32
• blog(“Implementing Recurrent Neural Net using chainer!”): http://t-satoshi.blogspot.jp/2015/06/implementing-recurrent-neural-net-using.html
• reddit:https://www.reddit.com/r/MachineLearning/comments/31rinf/150400941_a_simple_way_to_initialize_recurrent/
• reddit:https://www.reddit.com/r/MachineLearning/comments/32tgvw/has_anyone_been_able_to_reproduce_the_results_in/

Sequence Level Training with Recurrent Neural Networks

• arxiv: http://arxiv.org/abs/1511.06732
• notes: https://www.evernote.com/shard/s189/sh/ada01a82-70a9-48d4-985c-20492ab91e84/8da92be19e704996dc2b929473abed46

Papers

Generating Sequences With Recurrent Neural Networks

• arxiv: http://arxiv.org/abs/1308.0850
• github: https://github.com/hardmaru/write-rnn-tensorflow
• blog: http://blog.otoro.net/2015/12/12/handwriting-generation-demo-in-tensorflow/

DRAW: A Recurrent Neural Network For Image Generation

• arXiv: http://arxiv.org/abs/1502.04623
• github: https://github.com/vivanov879/draw
• github(Theano): https://github.com/jbornschein/draw
• github(Lasagne): https://github.com/skaae/lasagne-draw

Unsupervised Learning of Video Representations using LSTMs(ICML2015)

• project: http://www.cs.toronto.edu/~nitish/unsupervised_video/
• paper: http://arxiv.org/abs/1502.04681
• code: http://www.cs.toronto.edu/~nitish/unsupervised_video/unsup_video_lstm.tar.gz
• github: https://github.com/emansim/unsupervised-videos

LSTM: A Search Space Odyssey

• paper: http://arxiv.org/abs/1503.04069
• notes: https://www.evernote.com/shard/s189/sh/48da42c5-8106-4f0d-b835-c203466bfac4/50d7a3c9a961aefd937fae3eebc6f540
• blog(“Dissecting the LSTM”): https://medium.com/jim-fleming/implementing-lstm-a-search-space-odyssey-7d50c3bacf93#.crg8pztop
• github: https://github.com/jimfleming/lstm_search

Inferring Algorithmic Patterns with Stack-Augmented Recurrent Nets

• paper: http://arxiv.org/abs/1503.01007
• code: https://github.com/facebook/Stack-RNN

A Critical Review of Recurrent Neural Networks for Sequence Learning

• arXiv: http://arxiv.org/abs/1506.00019
• intro: “A rigorous & readable review on RNNs” 
  http://blog.terminal.com/a-thorough-and-readable-review-on-rnns/

Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks(Winner of MSCOCO image captioning challenge, 2015)

• arXiv: http://arxiv.org/abs/1506.03099

Visualizing and Understanding Recurrent Networks(Andrej Karpathy, Justin Johnson, Fei-Fei Li)

• paper: http://arxiv.org/abs/1506.02078
• slides:http://www.robots.ox.ac.uk/~seminars/seminars/Extra/2015_07_06_AndrejKarpathy.pdf

Grid Long Short-Term Memory

• arxiv: http://arxiv.org/abs/1507.01526
• github(Torch7): https://github.com/coreylynch/grid-lstm/

Depth-Gated LSTM

• paper: http://arxiv.org/abs/1508.03790
• code: GitHub(dglstm.h+dglstm.cc)

Deep Knowledge Tracing

• paper: https://web.stanford.edu/~cpiech/bio/papers/deepKnowledgeTracing.pdf
• github: https://github.com/chrispiech/DeepKnowledgeTracing

Top-down Tree Long Short-Term Memory Networks

• arxiv: http://arxiv.org/abs/1511.00060
• github: https://github.com/XingxingZhang/td-treelstm

Alternative structures for character-level RNNs(INRIA & Facebook AI Research)

• arXiv: http://arxiv.org/abs/1511.06303
• github: https://github.com/facebook/Conditional-character-based-RNN

Pixel Recurrent Neural Networks (Google DeepMind)

• arxiv: http://arxiv.org/abs/1601.06759
• notes(by Hugo Larochelle): https://www.evernote.com/shard/s189/sh/fdf61a28-f4b6-491b-bef1-f3e148185b18/aba21367d1b3730d9334ed91d3250848

Long Short-Term Memory-Networks for Machine Reading

• arxiv: http://arxiv.org/abs/1601.06733

Lipreading with Long Short-Term Memory

• arxiv: http://arxiv.org/abs/1601.08188

Associative Long Short-Term Memory

• arxiv: http://arxiv.org/abs/1602.03032

Representation of linguistic form and function in recurrent neural networks

• arxiv: http://arxiv.org/abs/1602.08952

Architectural Complexity Measures of Recurrent Neural Networks

• arxiv: http://arxiv.org/abs/1602.08210

Easy-First Dependency Parsing with Hierarchical Tree LSTMs

• arxiv: http://arxiv.org/abs/1603.00375

Training Input-Output Recurrent Neural Networks through Spectral Methods

• arxiv: http://arxiv.org/abs/1603.00954

Learn To Execute Programs

Learning to Execute

• arXiv: http://arxiv.org/abs/1410.4615
• github: https://github.com/wojciechz/learning_to_execute

Neural Programmer-Interpreters (Google DeepMind)

 

 

 

• arXiv: http://arxiv.org/abs/1511.06279
• project page: http://www-personal.umich.edu/~reedscot/iclr_project.html

A Programmer-Interpreter Neural Network Architecture for Prefrontal Cognitive Control

• paper:https://www.researchgate.net/publication/273912337_A_ProgrammerInterpreter_Neural_Network_Architecture_for_Prefrontal_Cognitive_Control

Convolutional RNN: an Enhanced Model for Extracting Features from Sequential Data

• arxiv: http://arxiv.org/abs/1602.05875

Attention Models

Recurrent Models of Visual Attention (Google DeepMind. NIPS2014)

• paper: http://arxiv.org/abs/1406.6247
• data: https://github.com/deepmind/mnist-cluttered
• code: https://github.com/Element-Research/rnn/blob/master/examples/recurrent-visual-attention.lua

Recurrent Model of Visual Attention(Google DeepMind)

• paper: http://arxiv.org/abs/1406.6247
• GitXiv: http://gitxiv.com/posts/ZEobCXSh23DE8a8mo/recurrent-models-of-visual-attention
• blog: http://torch.ch/blog/2015/09/21/rmva.html
• code: https://github.com/Element-Research/rnn/blob/master/scripts/evaluate-rva.lua

Show, Attend and Tell: Neural Image Caption Generation with Visual Attention

• paper: http://arxiv.org/abs/1502.03044
• code: https://github.com/kelvinxu/arctic-captions

A Neural Attention Model for Abstractive Sentence Summarization(EMNLP 2015. Facebook AI Research)

• arXiv: http://arxiv.org/abs/1509.00685
• github: https://github.com/facebook/NAMAS

Effective Approaches to Attention-based Neural Machine Translation(EMNLP2015)

• paper: http://nlp.stanford.edu/pubs/emnlp15_attn.pdf
• project: http://nlp.stanford.edu/projects/nmt/

Generating Images from Captions with Attention

• arxiv: http://arxiv.org/abs/1511.02793
• github: https://github.com/emansim/text2image
• demo: http://www.cs.toronto.edu/~emansim/cap2im.html

Attention and Memory in Deep Learning and NLP

• blog: http://www.wildml.com/2016/01/attention-and-memory-in-deep-learning-and-nlp/

Survey on the attention based RNN model and its applications in computer vision

• arxiv: http://arxiv.org/abs/1601.06823

Train RNN

Training Recurrent Neural Networks (PhD thesis)

• atuhor: Ilya Sutskever
• thesis: https://www.cs.utoronto.ca/~ilya/pubs/ilya_sutskever_phd_thesis.pdf

Deep learning for control using augmented Hessian-free optimization

• blog: https://studywolf.wordpress.com/2016/04/04/deep-learning-for-control-using-augmented-hessian-free-optimization/
• github: https://github.com/studywolf/blog/blob/master/train_AHF/train_hf.py

---

Hierarchical Conflict Propagation: Sequence Learning in a Recurrent Deep Neural Network

• arxiv: http://arxiv.org/abs/1602.08118

Recurrent Batch Normalization

• arxiv: http://arxiv.org/abs/1603.09025
• github: https://github.com/iassael/torch-bnlstm

Optimizing Performance of Recurrent Neural Networks on GPUs

• arxiv: http://arxiv.org/abs/1604.01946
• github: https://github.com/parallel-forall/code-samples/blob/master/posts/rnn/LSTM.cu

Codes

NeuralTalk (Deprecated): a Python+numpy project for learning Multimodal Recurrent Neural Networks that describe images with sentences

• github: https://github.com/karpathy/neuraltalk

NeuralTalk2: Efficient Image Captioning code in Torch, runs on GPU

• github: https://github.com/karpathy/neuraltalk2

char-rnn in Blocks

• github: https://github.com/johnarevalo/blocks-char-rnn

Project: pycaffe-recurrent

• code: https://github.com/kuprel/pycaffe-recurrent/

Using neural networks for password cracking

• blog: https://0day.work/using-neural-networks-for-password-cracking/
• github: https://github.com/gehaxelt/RNN-Passwords

Recurrent neural networks for decoding CAPTCHAS

• blog: https://deepmlblog.wordpress.com/2016/01/12/recurrent-neural-networks-for-decoding-captchas/
• demo: http://simplecaptcha.sourceforge.net/
• code: http://sourceforge.net/projects/simplecaptcha/

torch-rnn: Efficient, reusable RNNs and LSTMs for torch

• github: https://github.com/jcjohnson/torch-rnn

Deploying a model trained with GPU in Torch into JavaScript, for everyone to use

• blog: http://testuggine.ninja/blog/torch-conversion
• demo: http://testuggine.ninja/DRUMPF-9000/
• github: https://github.com/Darktex/char-rnn

LSTM implementation on Caffe

• github: https://github.com/junhyukoh/caffe-lstm

Blog

Survey on Attention-based Models Applied in NLP

http://yanran.li/peppypapers/2015/10/07/survey-attention-model-1.html

Survey on Advanced Attention-based Models

http://yanran.li/peppypapers/2015/10/07/survey-attention-model-2.html

Online Representation Learning in Recurrent Neural Language Models

http://www.marekrei.com/blog/online-representation-learning-in-recurrent-neural-language-models/

Fun with Recurrent Neural Nets: One More Dive into CNTK and TensorFlow

http://esciencegroup.com/2016/03/04/fun-with-recurrent-neural-nets-one-more-dive-into-cntk-and-tensorflow/

Materials to understand LSTM

https://medium.com/@shiyan/materials-to-understand-lstm-34387d6454c1#.4mt3bzoau

Understanding LSTM and its diagrams (★★★★★)

• blog: https://medium.com/@shiyan/understanding-lstm-and-its-diagrams-37e2f46f1714
• slides: https://github.com/shi-yan/FreeWill/blob/master/Docs/Diagrams/lstm_diagram.pptx

Persistent RNNs: 30 times faster RNN layers at small mini-batch sizes (Greg Diamos, Baidu Silicon Valley AI Lab)

http://svail.github.io/persistent_rnns/

All of Recurrent Neural Networks

https://medium.com/@jianqiangma/all-about-recurrent-neural-networks-9e5ae2936f6e#.q4s02elqg

Resources

Awesome Recurrent Neural Networks - A curated list of resources dedicated to RNN

• homepage: http://jiwonkim.org/awesome-rnn/
• github: https://github.com/kjw0612/awesome-rnn

Jürgen Schmidhuber’s page on Recurrent Neural Networks

http://people.idsia.ch/~juergen/rnn.html

Reading and Questions

Are there any Recurrent convolutional neural network network implementations out there ?

• reddit:https://www.reddit.com/r/MachineLearning/comments/4chu3y/are_there_any_recurrent_convolutional_neural/

 

« Reinforcement LearningSaliency Prediction »

 

 

Saliency Prediction

 Published: 09 Oct 2015  Category: deep_learning

This task involves predicting the salient regions of an image given by human eye fixations.

Large-scale optimization of hierarchical features for saliency prediction in natural images

• paper: http://coxlab.org/pdfs/cvpr2014_vig_saliency.pdf

Predicting Eye Fixations using Convolutional Neural Networks

• paper: http://www.escience.cn/system/file?fileId=72648

DeepFix: A Fully Convolutional Neural Network for predicting Human Eye Fixations

• arxiv: http://arxiv.org/abs/1510.02927

DeepSaliency: Multi-Task Deep Neural Network Model for Salient Object Detection

• arxiv: http://arxiv.org/abs/1510.05484

SuperCNN: A Superpixelwise Convolutional Neural Network for Salient Object Detection

• paper: www.shengfenghe.com/supercnn-a-superpixelwise-convolutional-neural-network-for-salient-object-detection.html

Shallow and Deep Convolutional Networks for Saliency Prediction

• arxiv: http://arxiv.org/abs/1603.00845
• github: https://github.com/imatge-upc/saliency-2016-cvpr

 

Scene Labeling

 Published: 09 Oct 2015  Category: deep_learning

Papers

Learning hierarchical features for scene labeling

• intro: “Their approach comprised of densely computing multi-scale CNN features for each pixel and aggregating them over image regions upon which they are classified. However, their methodstill required the post-processing step of generating over-segmented regions, like superpixels, for obtaining the final segmentation result. Additionally, the CNNs used for multi-scale feature learning were not very deep with only three convolution layers.”
• paper: http://yann.lecun.com/exdb/publis/pdf/farabet-pami-13.pdf

Indoor Semantic Segmentation using depth information

• arxiv: http://arxiv.org/abs/1301.3572

Multi-modal unsupervised feature learning for rgb-d scene labeling

• paper: http://www3.ntu.edu.sg/home/wanggang/WangECCV2014.pdf

Using neon for Scene Recognition: Mini-Places2

• intro: This is an implementation of the deep residual network used for Mini-Places2 as described in He et. al., “Deep Residual Learning for Image Recognition”.
• blog: http://www.nervanasys.com/using-neon-for-scene-recognition-mini-places2/
• github: https://github.com/hunterlang/mpmz

Attend, Infer, Repeat: Fast Scene Understanding with Generative Models

• arxiv: http://arxiv.org/abs/1603.08575

Challenges

Large-scale Scene Understanding Challenge

• homepage: http://lsun.cs.princeton.edu/