Keras R Accuracy History 보기

     batch_size <- 32

    # See ?dataset_cifar10 for more info

   cifar10 <- dataset_cifar10()

    x_train <- cifar10$train$x/255

    x_test <- cifar10$test$x/255

    bf_y<-cifar10$train$y    

    y_train <- to_categorical(cifar10$train$y, num_classes = 10)

    y_test <- to_categorical(cifar10$test$y, num_classes = 10)

    bf_y2<-cifar10$test$y

    

    # Defining Model ----------------------------------------------------------

    

    # Initialize sequential model

    model <- keras_model_sequential()

    

    model %>%

      

      # Start with hidden 2D convolutional layer being fed 32x32 pixel images

      layer_conv_2d(

        filter = 32, kernel_size = c(3,3), padding = "same", 

        input_shape = c(32, 32, 3)

      ) %>%

      layer_activation("relu") %>%

      

      # Second hidden layer

      layer_conv_2d(filter = 32, kernel_size = c(3,3)) %>%

      layer_activation("relu") %>%

      

      # Use max pooling

      layer_max_pooling_2d(pool_size = c(2,2)) %>%

      layer_dropout(0.25) %>%

      

      # 2 additional hidden 2D convolutional layers

      layer_conv_2d(filter = 32, kernel_size = c(3,3), padding = "same") %>%

      layer_activation("relu") %>%

      layer_conv_2d(filter = 32, kernel_size = c(3,3)) %>%

      layer_activation("relu") %>%

      

      # Use max pooling once more

      layer_max_pooling_2d(pool_size = c(2,2)) %>%

      layer_dropout(0.25) %>%

      

      # Flatten max filtered output into feature vector 

      # and feed into dense layer

      layer_flatten() %>%

      layer_dense(512) %>%

      layer_activation("relu") %>%

      layer_dropout(0.5) %>%

      

      # Outputs from dense layer are projected onto 10 unit output layer

      layer_dense(10) %>%

      layer_activation("softmax")

    

    opt <- optimizer_rmsprop(lr = 0.0001, decay = 1e-6)

    # model<-multi_gpu_model(model,gpus=2)

    model %>% compile(

      loss = "categorical_crossentropy",

      optimizer = opt,

      metrics = "accuracy"

    )

     

  

 ##################### 

  AccuracynHistory <- R6::R6Class("AccuracynHistory",

                                    inherit = KerasCallback, 

                                    

                                    public = list(

                                      val = NULL,tval=NULL,

                                      on_epoch_end = function(epoch,logs = list()) {

                                        preds <- do.call(cbind,list(model$predict(x_test)))

                                        value<-sum(apply(preds,1,which.max)-1==bf_y2)/length(bf_y2)

                                        self$val <- c(self$val,value)

                                        

                                        tpreds <- do.call(cbind,list(model$predict(x_train)))

                                        tvalue<-sum(apply(tpreds,1,which.max)-1==bf_y)/length(bf_y)

                                        self$tval <- c(self$tval,tvalue)

                                        

                                      }

                                    )

    )

     predictions <- AccuracynHistory$new()

   modd<- model %>% fit(

      x_train, y_train,

      batch_size = batch_size,

      epochs = 50,

     

      shuffle = TRUE,callbacks=predictions

    )  

   

   

   ts.plot(cbind(predictions$val,predictions$tval),col=c("red","blue"))