合页损失,铰链损失

Hinge Loss is a loss function used in Machine Learning for training classifiers. The hinge loss is a maximum margin classification loss function and a major part of the SVM algorithm.

铰链损失是机器学习中用于训练分类器的损失函数。 铰链损失是最大余量分类损失函数，是SVM算法的主要部分。

Hinge loss function is given by:

铰链损耗函数由下式给出：

Loss_H = max(0,(1-Y*y))

损耗_H = max(0，(1-Y * y))

Where, Y is the Label and

其中， Y是标签，

y = 𝜭.x

y = 𝜭.x

This is the general Hinge Loss function and in this tutorial, we are going to define a function for calculating the Hinge Loss for a multiple point (having one feature) with given 𝜭.

这是常规的Hinge Loss函数，在本教程中，我们将定义一个函数，用于计算给定𝜭的多点(具有一个特征)的Hinge Loss。

用于多点铰链损失的Python代码 (Python code for hinge loss for multiple points)

# Linear Algebra Learning Sequence# Hinge loss for Multiple Point import numpy as npdef hinge_loss_single(feature_vector, label, theta, theta_0):    ydash = label*(np.matmul(theta,feature_vector) + theta_0)    hinge = np.max([0.0, 1 - ydash*label])    return hingedef hinge_loss_full(feature_matrix, labels, theta, theta_0):    tothinge = 0    num = len(feature_matrix)    for i in range(num):        tothinge = tothinge + hinge_loss_single(feature_matrix[i], labels[i], theta, theta_0)            hinge = tothinge        return hingefeature_matrix = np.array([[2,2], [3,3], [7,0], [14,47]])theta = np.array([0.002,0.6])theta_0 = 0labels = np.array([[1], [-1], [1], [-1]])hingell = hinge_loss_full(feature_matrix, labels, theta, theta_0)print('Data point: ', feature_matrix)print('\n\nCorresponding Labels: ', labels)print('\n\n Hingle Loss for given data :', hingell)

Output:

输出：

Data point:  [[ 2  2] [ 3  3] [ 7  0] [14 47]]Corresponding Labels:  [[ 1] [-1] [ 1] [-1]] Hingle Loss for given data : [0.986]

翻译自:

合页损失,铰链损失