项目作者: purelyvivid

项目描述 :
Practice Deep Learning with Tensorflow, Keras, and pure Numpy, including BP algo, CNN, RNN, AotoEncoder, and Adversarial AotoEncoder.
高级语言: Python
项目地址: git://github.com/purelyvivid/DeepLearning_practice.git
创建时间: 2017-09-16T07:39:15Z
项目社区:https://github.com/purelyvivid/DeepLearning_practice

开源协议:

下载


DeepLearning_practice

1. NIN on cifar-10 with Keras

2. retrained VGG19 on cifar10 with Keras

3. BP alogrithm

  • [code]
  • A Deep Neural Network (DNN) model without toolkit for classification.
  • dataset: Street View House Number (SVHN)
  • loss: Cross-Entropy
  • opt: Stochastic Gradient Descent (SGD)
  • reg: N/A, L1, L2

4. DNN with Tensorflow

  • [code][4-1]
    [4-2]
  • A Deep Neural Network (DNN) model with tensorflow for classification.
  • dataset: MNIST
  • reg: L1, L2, Dropout

5. CNN with Tensorflow

  • [code][5-1]
    [5-2]
  • A Convolutional Neural Network(CNN) with tensorflow for classification.
  • dataset: cifar-10

6. RNN with Tensorflow

  • [code][6-1]
    [6-2]
  • A Recurrent Neural Network (RNN) with tensorflow for classification.
  • dataset: IMDB

7. Autoencoder with tensorflow

  • [code]
  • An Autoencoder with tensorflow.
  • dataset: MNIST with noise (MNIST-M was artificially created by using each MNIST digit as a binary mask and changing the colors of a background image.)

8. AAE with tensorflow

  • [code] [paper]
  • An Adversarial Autoencoder (AAE) with tensorflow.
  • dataset: MNIST