--- tags: ML --- 機器學習 入門(X 認識(O 03 === https://hackmd.io/NKAOz3geTqijBhR9r82RZw ### 笑話開場 神探狄仁傑向妻子抱怨道：「現在天道不彰，仁道不行，我名字之中有個仁字又有何用?」妻子道：「你跟我說這個幹嘛?」神探回答：「如果我是狄傑，你會愛我嗎?」 ### (簽到區) 幫我用留言功能簽到 # 複習 ### 監督式學習 1. 載入資料 >csv,xlxs... 2. 建立模型 >DNN,CNN... 4. 訓練模型 >次數,時間.... 5. 測試 >over fiting ### DNN 模組化輸出 # 補充 Dropout ```python= model.add(Dropout(0.2)) ``` ### back propragation # CNN 總圖 ![](https://2.bp.blogspot.com/-lo4ANmafiQM/WVpEeo68VPI/AAAAAAAAWxE/7FJBUGp-GrQFL5fUYEg6tQAJWXUHrtgWgCLcBGAs/s1600/3812_1.png) ## 概念 白板-> ## convolution ![](https://www.researchgate.net/profile/Chaim_Baskin/publication/318849314/figure/fig1/AS:614287726870532@1523469015098/Image-convolution-with-an-input-image-of-size-7-7-and-a-filter-kernel-of-size-3-3.png) filter 數量? 1 filter = 1 種特徵 ## max pool 降低資料大小 ![](https://www.google.com/url?sa=i&url=https%3A%2F%2Fpaperswithcode.com%2Fmethod%2Fmax-pooling&psig=AOvVaw1nplEqCwt7EKQlM8bsF51b&ust=1607003001838000&source=images&cd=vfe&ved=0CAIQjRxqFwoTCICM7Oq2r-0CFQAAAAAdAAAAABAI) ## Flatten 展開為一維，並接上NN >Flatten就是攤平之後，給你百八十個參數輸入，接上全連接層的腿 ## 統整 ![image alt](https://www.jefkine.com/assets/images/conv.png) ### back propragation 白板-> # 實戰 [機器學習02](https://hackmd.io/TAYAmalVTsKzzUrjLAZs_g) https://keras.io/zh/layers/convolutional/ ```python= def load_data():# (X_train, Y_train), (X_test, Y_test) = kr.datasets.mnist.load_data() x_train = X_train.reshape(60000, 1, 28, 28)/255 x_test = X_test.reshape(10000, 1, 28, 28)/255 y_train = np_utils.to_categorical(Y_train) y_test = np_utils.to_categorical(Y_test) return (x_train, y_train), (x_test, y_test) ``` ```python= model.add(Conv2D(filters=32, kernel_size=3, input_shape=(1, 28, 28), activation='relu', padding='same')) model.add(MaxPool2D(pool_size=2, data_format='channels_first')) model.add(Flatten()) ``` 原本 ```python= cv2_imshow(x_test[i]) ``` 改成 ```python= cv2_imshow((x_test[i]*255).reshape(28,28)) ``` filters:filter 的數量 kernel_size: kernel的大小 只設一個的話就是都一樣 activation:你知道的 padding:填充 ### 儲存模型 ```python= model.save('my_model.h5') ``` ### 載入模型 ```python= new_model=kr.models.load_model('my_model.h5') new_model.summary() los=model.evaluate(x_test,y_test) ``` ## 參考 https://blog.gtwang.org/programming/keras-save-and-load-model-tutorial/ https://medium.com/@CinnamonAITaiwan/%E6%B7%B1%E5%BA%A6%E5%AD%B8%E7%BF%92-cnn%E5%8E%9F%E7%90%86-keras%E5%AF%A6%E7%8F%BE-432fd9ea4935