This repository has been archived by the owner on Jan 15, 2021. It is now read-only.

caidwang / NN-and-DL-Notebook Public archive

Notifications You must be signed in to change notification settings
Fork 0
Star 1

Notebook of Neural Networks and Deep Learning, along with Reinforcement Learning.

1 star 0 forks Branches Tags Activity

Name		Name	Last commit message	Last commit date
Latest commit History 27 Commits
ReinforcementLearning		ReinforcementLearning
src		src
.gitignore		.gitignore
C1Week2.ipynb		C1Week2.ipynb
C1Week3.ipynb		C1Week3.ipynb
C2Week1.ipynb		C2Week1.ipynb
C2Week2.ipynb		C2Week2.ipynb
C2Week3.ipynb		C2Week3.ipynb
C5Week1.ipynb		C5Week1.ipynb
C5Week2.ipynb		C5Week2.ipynb
C5Week3.ipynb		C5Week3.ipynb
How to Apply Financial Aid for this Specialization.md		How to Apply Financial Aid for this Specialization.md
README.md		README.md

Repository files navigation

Deep Learning Specialization 学习笔记

吴恩达的Deep Learning Specialization 知识点整理, 内容暂时跳过了CNN部分, 建议与课程一起食用, 这部分作为复习时快速过知识点, 课程参考Coursera.

Course 1 Neural Networks and Deep Learning

Week2
- logistic回归 -- 二分类器
  - sigmoid函数
  - loss函数 && MSE和Cross Entropy方法对比
  - cost函数
  - 前向传播与反向传播
- numpy中的广播机制
- numpy的二分类器实现
Week3
- 浅层神经网络
- 神经网络的表示
- 正向传播和反向传播
- 激活函数 && sigmoid和relu, tanh的比较
- 梯度下降的学习方法
- 随机初始化

Course 2 Hyperparameter tuning, Regularization and Optimization

Week1
- train/Dev/Test Set
- 利用Bias和Variance诊断模型
- 正则化
  - L1/L2
  - Dropout
  - others
- 归一化
- 梯度爆炸和梯度消失
- Xavier参数初始化
Week2
- Mini-batch训练方法和优点
- 指数加权平均
  - Bias Correction
- Momentum SGD
- RMSprop
- Adam
- 学习率衰减
  - 神经网络优化中的局部最优问题
Week3
- 超参调整
  - 调整顺序(重要性排列)
  - 调整方法
- Batch Normalization
  - 计算方法
  - 机制有效性
  - BN在test过程的正确性
- TensorFlow基础
  - tensorflow基本操作
  - 使用tensorFlow训练全连接多分类器

Course3 Structuring Machine Learning Projects Todo!

Course5 Sequence Model

Week1
- RNN 基本结构
  - RNN的正向传播
  - BPTT反向传播
  - 梯度消失问题
  - 梯度爆炸/Clipping方法
  - RNN的numpy实现
- GRU
  - pytorch实现
- LSTM
  - pytorch实现
- 双向RNN
- Deep RNN
Week2 !todo
- 词嵌入
- Word2Vec && GloVe
Week3
- Seq2Seq
- Beam Search
- Attention Model
- Attention变种
  - Self Attention
  - Multi-head Attention

Reinforcement Learning

Others