This blog will continue to update this catalog, is my basic access to some of the current knowledge, we will continue to expand.
Artificial Intelligence Learning 1
- Environment, command 5 Anaconda
1.1. PIP 5
1.1.1. PIP install tensorflow 5
1.1.2. --Upgrade tensorflow 5 install updated PIP
1.1.3. PIP Show tensorflow 5
1.1.4. PIP install --upgrade --ignore- tensorflow Installed.. 6
1.2. context switch. 6
1.2.1. tensorflow the activate. 6
1.2.2. Conda the deactivate. 6
1.3. channels 6
1.3.1. restore default channels 6 - tensorflow 6
2.1. session 6 - 7 Spyder
3.1. Shortcuts 7
3.1.1. Note 7 - Protobuf 7
4.1.1. Installation 7
4.1.2 8 using - Mathematical. 9
5.1. The ROC curve, the Measure-Fl, IOU. 9
5.1.1. The ROC curve. 9
5.1.2. Fl-the Measure 10
5.1.3. The IOU 10
5.2. 10 Robustness
5.3. Perceptron 11
5.4. Gradient 12
5.4 .1. linear regression 12
5.4.2 least mean squares method (Least on Mean Squares) 13 is
5.4.3. gradient descent the GD (gradient Descent) 13 is
5.4.4. batch gradient descent algorithm (BGD) 14
5.4.5. stochastic gradient descent (Stochastic Gradient Descent, the SGD) 15
. AdaGrad 16 5.4.6
. 5.4.7. 17 Adadelta
5.4.8 algorithm for the momentum - momentum algorithm. 17
. 5.4.9 RMSprop algorithm 20 is
. Adam 5.4.10 algorithm 21
5.5 21 activation function.
5.5 .1. Sigmoid activation function 21 is
5.5.2. RELU activation function 22
5.6. 22 function the Softmax
5.6.1. classification 23 is
5.6.2. SoftMax regression model 23 is
5.6.3. SoftMax operation 24
5.6.4 single sample classification calculation expression vector 25
5.6.5. Classification of small quantities of sample calculation expression vector 26
5.6.6 cross entropy loss function 27
5.6.7. Prediction and Evaluation Model 28
5.6.8. Summary 28
5.7. 28 Dropout
5.8. BN (BatchNorm) 29
5.9. hot encoded (One-hot) 34
5.10. embedding embedding that word 34
5.11. dimensionality reduction algorithm UMAP, the SNE T-34
5.12. on BP back propagation algorithm (backpropagation) 38
. About 38 5.12.1
5.12.2 step a: propagating forward 40
5.12.3 step two: backpropagation 41
5.12.4 46 are summarized.
5.12.5 gradient disappears, gradient 47 explosion. - Image 48
6.1. MAP 48
6.2. Detection of the target IOU 48
6.3. Receptive field theory and the effective receptive field 49
6.4. Anchor 50
6.5. Downsampling on a sample 51
6.5.1. Conventional interpolation method 52 is
6.6. FCN- image semantic segmentation 63
6.6.1. the FCN structure 63
6.6.2 upsampler 65
6.7. semantic segmentation image the U-Net-66 - Image model 67
7.1. YOLO3- object detection 67 - Model 68
8.1. AlexNet (convolutional neural network) 68
8.1.1. AlexNet features 68
8.1.2 respective local normalization 70
8.2. VGGNet (convolutional neural network) 70
8.2.1. VGGNet Structure 70
8.3. The ResNet ( residual network) 72
8.4. RNN (Recurrent neural network) 74
8.4.1. RNN model structure 74
8.4.2. RNN backpropagation 78
8.4.3. LSTM and GRU (RNN some improved algorithm) 81
8.5. the SDD (Single Shot MultiBox Detector) - Object detection 85
8.6 FPN (the feature pyramid Network) -. features 86 pyramids
8.6.1 Gaussian pyramid 87.
8.6.2 feature of the pyramid 88.
8.6.3 FPN 89.
8.7 Faster RCNN \ RPN -. image classification 92
8.8. Inception- convolutional architecture 93
8.8.1. the Inception V1 94
8.8.2. the Inception V2 94
8.8.3. the Inception V3 95
8.8.4. 96 the Inception V4
8.8.5. the ResNet the Inception-96
8.9. MobileNet 97
8.9.1. Separable convolution depth 97
8.9.2. MobileNet Vl 98
8.9.3. MobileNet V2 101
8.9.4. MoblieNet 104 V3
8.10. MTCNN- multitasking convolutional neural network 107
8.10.1. What MTCNN 107
8.10.2. constructing an image pyramid 108
8.10.3. P-Net 108
8.10.4. R & lt Net-109
8.10.5. O-Net 110
8.10.6. Thought integration architecture and system 111
8.11. CTPN- text recognition 113
8.12. CRNN- end identification 113
8.13. input and output matching CTC-113 - Reinforcement Learning 113
9.1. Q-Learning 114
9.2. 114 on Sarsa
9.3. DQN (the Network Deep Q) 115
9.4. Gradient 115 the Policy - 数据集dataset 116
10.1.1. ICDAR 116
10.1.2. the Almighty dataset 117 - 117 the NLP
11.1. SLING- semantic parser 117
