Deep Learning related code examples. (Including the exercises of UFLDL)
Dependence: Numpy Scipy Matplotlib Sklearn.
TIP: The first seven are the UFLDL exercises implement in python.
####SparseAutoEncoder
- Sparse autoencoder.
- 稀疏自编码
####PCA_Whitening
- PCA / Whitening based on PCA / ZCA Whitening based on PCA.
- 主成分分析 / PCA白化 / ZCA白化
####SoftmaxRegression
- SoftMax Regression, generalization of logistic regression.
- SoftMax回归
####SelfTaughtLearning
- Extract features by Autoencoder, then make SoftMax Regression.
- 自编码 --> SoftMax
####DeepNetworks
- Extract features by two hidden layers Autoencoder (and Fine-tuning with the labeled data), then make SoftMax Regression.
- 深度多层自编码 (--> 微调 ) --> SoftMax
####LinearDecoder
- ZCAWhiten the data and get the parameters of Autoencoder which has a linear decoder, then save the parameters.
- ZCA白化 --> 线性解码的稀疏自编码 --> 获取预处理参数
####ConvolutionPooling
- Extract features with the parameters above and convolve/pool the features, then make SoftMax Regression.
- 应用预处理参数 --> 卷积/池化 --> SoftMax
-
Reference
-
TODOS:
- Advanced Topics of UFLDL
- Sparse Coding
- ICA Style Models
- Other deep learning related code.
- Advanced Topics of UFLDL