svmlin

import numpy as np
# might need to add path to mingw-w64/bin for cvxopt to work
# import os
# os.environ["PATH"] += os.pathsep + ...
import cvxopt




def svmlin(X, t, C):
    # Linear SVM Classifier
    #
    # INPUT:
    # X        : the dataset                  (num_samples x dim)
    # t        : labeling                     (num_samples x 1)
    # C        : penalty factor for slack variables (scalar)
    #
    # OUTPUT:
    # alpha    : output of quadprog function  (num_samples x 1)
    # sv       : support vectors (boolean)    (1 x num_samples)
    # w        : parameters of the classifier (1 x dim)
    # b        : bias of the classifier       (scalar)
    # result   : result of classification     (1 x num_samples)
    # slack    : points inside the margin (boolean)   (1 x num_samples)


    #####Insert your code here for subtask 2a#####

    dim = X.shape(1)
    print(dim)
    N = t.size
    print("X",X.shape)
    H = np.zeros((N,N))
    for i in range(0, N):
        for j in range(0, N):
            H[i,j] = t[i]*t[j]*np.dot(X[i]*X[j])
    n = H.shape[1]
    print("n:", H.shape)


    P = cvxopt.matrix(H)
    print("P", P )
    q = cvxopt.matrix((-1)*np.ones(N))
    G = np.vstack([-np.eye(n), np.eye(n)]).astype('double')
    #G = G[:2,:]
    G = cvxopt.matrix(G)
    print("G", G)
    A = t.astype('double')[np.newaxis]
    #A = np.transpose(A)

    #A = np.matrix.transpose(A)
    print(np.size(A))
    print("wtf")
    A = cvxopt.matrix(A)
    print("A", A)

    b = cvxopt.matrix(np.zeros(1))
    z = np.zeros(N)
    c = C *np.ones(N)
    h = (np.hstack([z,c])).astype('double')
    print("h", h.shape, h)

    h = cvxopt.matrix(h)


    print(np.size(b))
    print(b)
    a = cvxopt.solvers.qp(P, q, G, h, A, b)
    print(a["z"])

    print("d")
    alpha = a["x"]
    sv = [None]*N


    for i in range(0,N):
        if a["x"][i]==0:
            sv[i] = False
        else:
            sv[i] = True

    w= np.zeros((1,))
    for i in range(0,N):
        w[0] = w[0] + alpha[i]*t[i]*X[i,0]
        w[1] = w[1] + alpha[i]*t[i]*X[i,1]

    print("w",w)

    b,nm, b1 = 0
    for i in range(0,N):
        if(alpha[i]<C and alpha[i]>0):
            nm = nm+1

    for i in range(0,N):
        b1 = b1 + alpha[i]*t[i]*X[i,0]
        b1 = b1 + alpha[i]*t[i]*X[i,1]

    b1 = b1
    result = 1
    slack = 1



    return alpha, sv, w, b, result, slack