main.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import numpy as np
import matplotlib.pyplot as plt
import methods as meth
import functions as f
import plotting as plot
import sys



'''Constants:'''
m = 2000 # number of points
n = 2 # dimensions
area = 2.0
x_length = int(n*(n+1)/2)+n
prob = 0.02
min_rec, max_rec = 1, 4


lambda_low = 0.1
lambda_high = 10

# Create a feasible starting point
x_initial = np.zeros(x_length)
x_initial[0], x_initial[2] = (lambda_low+lambda_high)/2, (lambda_low+lambda_high)/2
x_initial[1] = np.sqrt((x_initial[0]*x_initial[2]-lambda_low**2)/2)
print("x_initial", x_initial)

if np.amin(f.c_function(x_initial, lambda_low, lambda_high)) < 1e-20:
    sys.exit("Illegal starting point")

if __name__ == "__main__":
    # Define the classifying ellipse.

    #A_classification = [[1, 0], [0, 1]]  # Circle
    #A_classification = [[0.7, 0.5], [0.5, 0.9]]  # Ellipse
    #A_classification = [[0.0001, 10], [10, 0.0001]]  # Hyperbola
    #A_classification = [[0.00001, 50], [50, 0.00001]]  # Hyperbola closer to asymptotes
    A_classification = [[0.1, 10], [10, 0.1]]  # A third hyperbola
    b_classification=[0, 0]

    #Create z-list. Comment in the preferred.
    z_list = f.construct_z_elliptic(n, m, A_classification, b_classification, area,newdataset=True)
    #z_list = plot.classify_misclassification(m, n, area, prob)

    solution = meth.primalBarrier(f.P, f.dP, z_list, n, x_initial, lambda_low, lambda_high)

    # The rest here is for plotting
    A_solution,b_solution=f.construct_A_and_b(n, solution)
    plot.plot_z_points(z_list, m)
    X, Y, Z = plot.make_ellipse(A_solution, b_solution, area, plot.eval_func_model_2D)
    plot.plot_dataset_2d(X, Y, Z)
    print("\neigenvalues of classification matrix\n", np.linalg.eigvals(A_classification))
    plt.show()