The points on one side are negative, The point result score on the other side is a positive example It is also an algorithm to solve binary classification problem by finding hyperplane The function distance is the relative distance between two points （SVM Using the functional distance ） Geometric distance is a formula to find the distance from a point to a plane Geometric distance and functional distance Logical regression is x And model θ Multiply and pass sigmoid Function maps the result to 0-1 Probability between, In this way, the probability can be multiplied by the loss function obtained by the maximum likelihood estimation.
The perceptron is x And model θ Multiply, If it is greater than 0, That is to say, they are classified into positive classes, If it is less than 0, Is classified as negative, So there is no probabilistic meaning, The perceptron is expected to minimize the sum of distances from all samples with classification errors to the hyperplane, The algorithm of perceptron is to find out the wrong samples, Find the distance from all the wrong samples to the hyperplane as small as possible. Before integrated learning ,SVM It is a popular algorithm in the field of classification. The essence of logistic regression is also a linear classifier, But it can do nonlinear classification （ L d ）.Ĥ. In high dimensional space m Samples as support vectors, Supporting a hyperplane, This hyperplane is the boundary. Perceptron requires data set to be linearly separableĢ. Support vector machine SVM It is an algorithm for linear classification, It can also do nonlinear classification （ L d ） Extension of perceptron algorithm model.
0 kommentar(er)
