Skip to content

Commit

Permalink
Update readme.md
Browse files Browse the repository at this point in the history
  • Loading branch information
Ewenwan authored May 27, 2018
1 parent 4c51b7d commit 3c79c05
Showing 1 changed file with 106 additions and 0 deletions.
106 changes: 106 additions & 0 deletions deepLearning/readme.md
Original file line number Diff line number Diff line change
Expand Up @@ -8,3 +8,109 @@
- 深度信念网络Deep Belief Network (DBN) [dbn.py](dbn.py)

Note: the project aims at imitating the well-implemented algorithms in [Deep Learning Tutorials](http://www.deeplearning.net/tutorial/) (coded by [Theano](http://deeplearning.net/software/theano/index.html)).


[吴恩达老师的深度学习课程笔记及资源](https://github.com/fengdu78/deeplearning_ai_books)

## 1. 神经网络的编程基础(Basics of Neural Network programming)
### 1.1 二分类(Binary Classification)
### 1.2 逻辑回归(Logistic Regression)
### 1.3 逻辑回归的代价函数(Logistic Regression Cost Function)
### 1.4 梯度下降(Gradient Descent)
### 1.5 导数(Derivatives)
### 1.6 计算图(Computation Graph)
### 1.7 逻辑回归的梯度下降(Logistic Regression Gradient Descent)
### 1.8 梯度下降的例子(Gradient Descent on m Examples)
### 1.9 向量化(Vectorization)
### 1.10 损失函数详解(cost/Loss function)softmax 交叉熵 Forc Loss

## 2. 浅层神经网络(Shallow neural networks)
### 2.1 计算一个神经网络的输出 (Computing a Neural Network's output)
### 2.2 多样本向量化(Vectorizing across multiple examples)
### 2.3 激活函数(Activation functions)
### 2.4 神经网络的梯度下降(Gradient descent for neural networks)
### 2.5 直观理解反向传播(Backpropagation intuition)
### 2.6 随机初始化(Random+Initialization)
### 2.7

## 2. 深层神经网络(Deep Neural Networks)
### 3.1 前向传播和反向传播(Forward and backward propagation)
### 3.2 深层网络中的前向和反向传播(Forward propagation in a Deep Network)
### 3.3 核对矩阵的维数(Getting your matrix dimensions right)
### 3.4 参数VS超参数(Parameters vs Hyperparameters)
### 3.5 深度学习和大脑的关联性(What does this have to do with the brain?)

## 3. 改善深层神经网络:超参数调试、正则化以及优化
### 3.1 数据集 训练测试验证 均值方差
### 3.2 正则化 L1 L2 Dropout 随机失活 正则化 Regularization
### 3.3 标准化输入(Normalizing inputs) BN GN
### 3.4 梯度消失/梯度爆炸(Vanishing / Exploding gradients)
### 3.5 神经网络的权重初始化(Weight Initialization for Deep NetworksVanishing /Exploding gradients)
### 3.6 梯度的数值逼近(Numerical approximation of gradients)& 梯度检验(Gradient checking)

## 4. 优化算法 (Optimization algorithms)
### 4.1 Mini-batch 梯度下降(Mini-batch gradient descent)
### 4.2 指数加权平均(Exponentially weighted averages)
### 4.3 momentum梯度下降(Gradient descent with momentum)
### 4.4 RMSprop——root mean square prop(RMSprop)
### 4.5 Adam优化算法(Adam optimization algorithm)
### 4.6 学习率衰减(Learning rate decay)
### 4.7 局部最优问题(The problem of local optima)



## 5. 卷积神经网络(Convolutional Neural Networks)
### 5.1 卷积层 普通卷积 DW逐通道卷积 分组卷积
### 5.2 池化层 最大值池化 均值池化
### 5.3 全连接层
### 5.4 Padding填充 步长Stride
### 5.5 经典网络(Classic networks)
### 5.6 残差网络(Residual Networks (ResNets))
### 5.7 网络中的网络以及 1×1 卷积(Network in Network and 1×1 convolutions)通道重排ChannelShuffle
### 5.8 Inception Xception ResXt DenseNet DectNet 网络
### 5.9 特征金字塔 FPN
### 5.10 目标检测OD 边框预测BBP 交并比IOU 非极大值抑制NMS 预设边框Anchor Boxes
### 5.11 YOLO v1 v2 v3 算法
### 5.12 轻量化模型 MobileNet SqueezeNet ShuffleNet

## 6. 人脸识别和神经风格转换(Special applications: Face recognition &Neural style transfer)
### 6.1 人脸识别
### 6.2 神经风格转换 neural style transfer
### 6.3 深度卷积网络学习 deep ConvNets learning
### 6.4 代价函数(Cost function) 内容代价函数(Content cost function) 风格代价函数(Style cost function)
### 6.5 一维到三维推广(1D and 3D generalizations of models)

## 7. 序列模型(Sequence Models)
### 7.1 循环神经网络模型 RNN(Recurrent Neural Network Model)
#### 7.1.1 通过时间的反向传播(Backpropagation through time)
#### 7.1.2 不同类型的循环神经网络(Different types of RNNs)
#### 7.1.3 语言模型和序列生成(Language model and sequence generation)
#### 7.1.4 对新序列采样(Sampling novel sequences)
#### 7.1.5 循环神经网络的梯度消失(Vanishing gradients with RNNs)
### 7.2 GRU单元(Gated Recurrent Unit(GRU))
### 7.3 长短期记忆(LSTM(long short term memory)unit)
### 7.4 双向循环神经网络(Bidirectional RNN)
### 7.5 深层循环神经网络(Deep RNNs)

## 8. 自然语言处理与词嵌入(Natural Language Processing and Word Embeddings)
### 8.0 语音识别的传统方法 HMM隐马尔可夫模型 GMM高斯混合模型
### 8.1 词汇表征(Word Representation)
### 8.2 使用词嵌入(Using Word Embeddings) 词嵌入的特性(Properties of Word Embeddings) 词嵌入除偏(Debiasing Word Embeddings)
### 8.3 嵌入矩阵(Embedding Matrix)
### 8.4 Word2Vec
### 8.5 负采样(Negative Sampling) GloVe 词向量(GloVe Word Vectors)
### 8.6 情绪分类(Sentiment Classification)
### 8.7 集束搜索(Beam Search)
### 8.8 注意力模型(Attention Model)
### 8.9 语音识别(Speech recognition)
### 9.10 触发字检测(Trigger Word Detection)

## 9. 视频分析 人体行为识别
### 9.1 传统方法 视觉词袋模型BoVW FV高斯混合建模
### 9.2 双流模型
### 9.3 3D卷积模型
### 9.4 双流+3D卷积
### 9.5 LSTM模型
###
###

0 comments on commit 3c79c05

Please sign in to comment.