Chapter 8: Support Vector Machines Flashcards
describe the idea of support vector machines
build a geometric view of separating data points
how do we get the plane in a linear model
set boundary equal to 0
i.e.
w1x1 + w2x2 + … + b = 0
how do we calculate the distance between a point and the hyperplane
|| w || 2
how do we assign classes in a support vector machine
based on whether the distance to the hyperplane is +ve or -ve
how do we calculate the distance from the origin to the hyperplane
b / ||w||2
how do we create two parallel hyperplanes
set equal +1 and -1
what is the distance between the two parallel hyperplanes
2 / || w ||2
what is the separation margin
the region bounded by two hyperplanes
what is an optimal hyperplane
separates the data points with the widest margin
what are support vectors
points that are on the line, The most difficult points to classify
what is done in the process of training an SVM
optimise the hyperplane
what does the lagrangian method do
a quadratic programming problem that solves the dual problem
what is the dual problem
minimising || w ||2 while ensuring all points are outside of the margin
what is a hard margin SVM
no points inside the margin
what is a soft margin SVM
points allowed inside the hyperplane and on the wrong sides
what is a slack variable
relaxes the constrains to allow points into the hyperplane
how does a soft margin SVM change the optimisation problem
also have to minimise the number of slack variables
what are support vectors in a soft margin SVM
elements within the margin
elements on the hyperplanes
elements on the wrong sides of the plane
how do we support non linear data patterns in SVM
kernel method
give two kernel methods
gaussian
polynomial
where is the kernel method added for a non linear SVM
in the dual problem
what is the decision function
how we calculate the optimal values for the multiplies using the results of the dual problem
how do we calculate the bias
using the slack variables
