Advantages: Data for the number of small samples and high dimensional feature
Goal: to achieve the purpose of binary classification
Select Hyperplane basis:
Unable to find other methods to draw the greater the distance between the two dashed lines
Its optimal hyperplane to the nearest point of the two types of data at the same distance
Spacing problem
Hard interval: may appear over-fitting phenomenon
Soft Margin: Allows training has some error
Linearly inseparable problem can be solved
L-dimensional mapping of high-dimensional space to solve the problem
Kernel:
Victoria is from low to high-dimensional mapping function
The most common core is: a Gaussian kernel
Combat:
from sklearn Import SVM # Import SVM bag X- = [[0,0], [2,2 &], [3,3], [4,4 &]] # of training data Y = [1,2,3,4 ] CLF svm.SVC = (Kernel = " RBF " , Gamma = ' Auto ' ) # initialization using radial basis classifier clf.fit (X-, Y) # training T = [[2,1], [0,1 ]] # test set Print (clf.predict (T)) Print (clf.decision_function (T)) # returns to the test data set from the hyperplane
Concern is decision_function () function call codes Finally, it returns the data set from the hyperplane expressed by the positive and negative side of the hyperplane in which, as the absolute value of the distance, the reliability of the classification higher.