support_vector_machines

support_vector_machines, looks at how support vector machines can classify data.

The notes:

• support_vector_machines.pdf
• support_vector_machines.tex

Scripts:

• cvxopt_example.py, uses cvxopt() to solve an example quadratic programming problem.
• generator_perceptron.py, uses gradient descent to solve the perceptron problem for a set of generators to be classified as "failing(0)" or "working(1)".
• jet_svm.py, uses cvxopt() to determine the SVM for the jet data.
• perceptron_gradient.py, uses gradient descent to solve a perceptron problem.

Text files:

• generator_data.txt, 56 records: index, rpm, vibration, fail(0)/active(1)
• generator_perceptron.txt, output from generator_percepton.py.
• gopher_data.txt, 50 records: skull width, skull length, golden(-1)/hopping(1).
• jet_data.txt, 56 records: index, rpm, vibration, fail(-1)/working(1)

Images:

• generator_data.png, plots rpm and vib for generators which are failing or working.
• generator_perceptron.png, plots rpm and vib for generators which are failing or working, and displays the perceptron dividing line.
• jet_svm.png, the data and SVM separating hyperplane for the jet data.
• separating_plane.png
• svm.png, illustrates two lines that could separate data.