博主水平灰常有限,期望能与對、Deep Learning 感兴趣的朋友一起学习、交流、探讨与分享~~

1. Neural Networks / Deep Learning

No. Neural Networks Neural Networks and Deep Learning
1 Introduction to Deep Learning 了解现今深度学习在哪里应用、如何应用
2 Neural Networks Basics logistic 损失函数、导数、计算图、m个样本的梯度下降、向量化
3 Shallow Neural Networks NN Representation、向量化、Activation functionsRandom init、Python 实现 NN
4 Deep Neural Networks 深网的前向传播、核对矩阵维数、反向传播、参数VS超参数、Python 实现 NN

对于中间层来说, 往往是 ReLU 的效果最好.
虽然 z < 0, 的时候,斜率为0, 但在实践中,有足够多的隐藏单元 令 z > 0, 对大多数训练样本来说是很快的.

so the one place you might use as linear activation function others usually in the output layer.

2. Improving Deep Neural Networks

No. Improving Improving Deep Neural Networks
5 Deep Learning Action Train/Dev/Test、Bias/Variance、L1 L2 Dropout、梯度消失\爆炸、Weight init、Gradient checking
6 Optimization mini-batch、指数加权平均-偏差修正、Momentum、RMSprop、Adam、α decay、局部优
7 超参数调试、Batch Hyperparameter、Normalizing Activations、Batch Norm [Fitting NN]、Softmax

3. Structured Machine Learning

No. Structured Structured Machine Learning Projects
8 ML Strategy 1 正交化、Satisficing and optimizing metrics、Train/dev/test 改变、可避免偏差、人的表现
9 ML Strategy 2 误差分析、标注错误数据、定位数据不匹配偏差与方差、迁移学习、多任务学习、端到端学习

4. Convolutional Neural Networks

No. CNN Convolutional Neural Networks
10 Convolutional Neural Networks Edge detection、Padding、Strided convolutions、Convolutions Over Volumes、Pooling
11   Deep CNN   Classic Nets、ResNets、1×1 convolutions、Inception、Transfer Learning、Data augmentation
12 Object detection Object Localization、Landmark Detection、Sliding Windows、Bounding Box Predictions、Intersection Over Union、Non-max Suppression、Anchor Boxes、YOLO
13 Face recognition One-Shot、Siamese、Triplet Loss、Face Verification、deep ConvNets learning?

LeNet-5、AlexNet、VGG、ResNet (有152层)、Inception。 目标定位、特征点检测、Bounding Box预测、Anchor Boxes

5. Sequence Models

No. Sequence Models Sequence Models
14 Recurrent Sequence Models Notation、RNN、Vanishing gradients、GRU、LSTM、BRNN、Deep RNNs
15 NLP & Word Embeddings Matrix、Word2Vec、Negative Sampling、GloVe、Debiasing Word Embeddings
16 Sequence Models & Attention Greedy Search、Beam Search、Error analysis on beam search、Attention

Contents

  1. 1. Neural Networks / Deep Learning
  2. 2. Improving Deep Neural Networks
  3. 3. Structured Machine Learning
  4. 4. Convolutional Neural Networks
  5. 5. Sequence Models
  6. Friends link