开始学习深度学习和循环神经网络Some starting points for deep learning and RNNs

Bengio,  LeCun,  Jordan,  Hinton,  Schmidhuber,  Ng,  de Freitas and  OpenAI have done reddit AMA's.  These are nice places to start to get a Zeitgeist of the field.

 

Hinton and  Ng lectures at  Coursera,  UFLDL,  CS224d and  CS231n at Stanford, the  deep learning course at  Udacity, and the  summer school at  IPAM have excellent tutorials, video lectures and programming exercises that should help you get started.

 

The online book by  Nielsen, notes for  CS231n, and blogs by  Karpathy,  Olah and  Britz have clear explanations of MLPs, CNNs and RNNs.  The tutorials at  UFLDL and  deeplearning.net give equations and code. The encyclopaedic book by  Goodfellow et al. is a good place to dive into details.  I have a  draft book in progress.

 

Theano,  Torch,  Caffe,  ConvNet,  TensorFlow,  MXNet,  CNTK,  Veles,  CGT,  Neon,  Chainer,  Blocks and  Fuel,  Keras,  Lasagne,  Mocha.jl,  Deeplearning4j,  DeepLearnToolbox,  Currennt,  Project Oxford,  Autograd ( for Torch),  Warp-CTC are some of the many deep learning software libraries and frameworks introduced in the last 10 years.   convnet-benchmarks and  deepframeworks compare the performance of many existing packages. I am working on developing an alternative,  Knet.jl, written in  Julia supporting CNNs and RNNs on GPUs and supporting easy development of original architectures.  More software can be found at  deeplearning.net.

Deeplearning.net and homepages of  Bengio,  Schmidhuber have further information, background and links.

from: http://www.denizyuret.com/2014/11/some-starting-points-for-deep-learning.html