Pegasos SVM Step

Problem

Take one Pegasos stochastic subgradient step for an SVM.

First shrink w by the factor (1 - eta * lam). If y * (w dot x) < 1, also add eta * y * x. Return the new weight vector.

Hint

The loss is the hinge function max(0, 1 - y * (w dot x)).

case 1

in: [[0,0],[1,2],1,0.1,0.1]

expected: [0.1,0.2]

case 2

in: [[1,1],[1,1],1,0.1,0.5]

expected: [0.95,0.95]

case 3

in: [[0.5,-0.5],[1,2],-1,0.2,0.1]

expected: [0.39,-0.69]

Starter codePython

def pegasos_step(w, x, y, lam, eta):
    # Your code here
    pass

3 test cases · run your code against them in the editor