我正在使用package de.freakyonline.ucone;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;
import java.net.Socket;
import javafx.application.Application;
import static javafx.application.Application.launch;
import javafx.fxml.FXMLLoader;
import javafx.scene.Parent;
import javafx.scene.Scene;
import javafx.stage.Stage;
public class MainApp extends Application {
String ver = "v0.1-SNAPSHOT";
ObjectOutputStream out;
ObjectInputStream in;
@Override
public void start(Stage stage) throws Exception {
Parent root = FXMLLoader.load(getClass().getResource("/fxml/Scene.fxml"));
// Connect to Server
try {
Socket sock = new Socket("unitycraft.de", 2009);
out = new ObjectOutputStream(sock.getOutputStream());
in = new ObjectInputStream(sock.getInputStream());
// Listen for remote stuff comming in ...
Thread remote = new Thread(new RemoteReader(in,out,sock));
remote.start();
} catch (Exception ex) { ex.printStackTrace(); }
Scene scene = new Scene(root);
scene.getStylesheets().add("/styles/Styles.css");
stage.setTitle("UCOne - The UnityCraft Staff Tool " + ver);
stage.setScene(scene);
stage.show();
}
/**
* The main() method is ignored in correctly deployed JavaFX application.
* main() serves only as fallback in case the application can not be
* launched through deployment artifacts, e.g., in IDEs with limited FX
* support. NetBeans ignores main().
*
* @param args the command line arguments
*/
public static void main(String[] args) {
launch(args);
}
}
（package de.freakyonline.ucone;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;
import java.net.Socket;
/**
*
* @author uwe
*/
public class RemoteReader implements Runnable {
Object obj = null;
ObjectInputStream in;
ObjectOutputStream out;
Socket sock;
public RemoteReader (ObjectInputStream in, ObjectOutputStream out, Socket sock) {
this.in = in;
this.out = out;
this.sock = sock;
}
public void run() {
try {
while((obj=in.readObject()) != null)
System.out.println("Got object from server ...");
} catch (Exception ex) { ex.printStackTrace(); }
}
}
后端）进行二进制分类，并且我有大约76％的精确度和70％的召回率。现在我想尝试使用决策阈值。据我所知Keras使用决策阈值0.5。 Tensorflow是否有办法使用自定义阈值来确定决策精度和召回？
感谢您的时间！
答案 0 :(得分：18)
创建这样的自定义指标：
编辑感谢@Marcin ：创建以threshold_value为参数返回所需指标的函数
def precision_threshold(threshold=0.5):
def precision(y_true, y_pred):
"""Precision metric.
Computes the precision over the whole batch using threshold_value.
"""
threshold_value = threshold
# Adaptation of the "round()" used before to get the predictions. Clipping to make sure that the predicted raw values are between 0 and 1.
y_pred = K.cast(K.greater(K.clip(y_pred, 0, 1), threshold_value), K.floatx())
# Compute the number of true positives. Rounding in prevention to make sure we have an integer.
true_positives = K.round(K.sum(K.clip(y_true * y_pred, 0, 1)))
# count the predicted positives
predicted_positives = K.sum(y_pred)
# Get the precision ratio
precision_ratio = true_positives / (predicted_positives + K.epsilon())
return precision_ratio
return precision
def recall_threshold(threshold = 0.5):
def recall(y_true, y_pred):
"""Recall metric.
Computes the recall over the whole batch using threshold_value.
"""
threshold_value = threshold
# Adaptation of the "round()" used before to get the predictions. Clipping to make sure that the predicted raw values are between 0 and 1.
y_pred = K.cast(K.greater(K.clip(y_pred, 0, 1), threshold_value), K.floatx())
# Compute the number of true positives. Rounding in prevention to make sure we have an integer.
true_positives = K.round(K.sum(K.clip(y_true * y_pred, 0, 1)))
# Compute the number of positive targets.
possible_positives = K.sum(K.clip(y_true, 0, 1))
recall_ratio = true_positives / (possible_positives + K.epsilon())
return recall_ratio
return recall
现在您可以在
中使用它们model.compile(..., metrics = [precision_threshold(0.1), precision_threshold(0.2),precision_threshold(0.8), recall_threshold(0.2,...)])
我希望这会有所帮助：）