grid_search_1d.py

from klee_minty import *
from solver import *
import pandas as pd
from tabulate import tabulate
import pickle

dimension_grid = [2, 4, 6]#$, 10, 14, 18, 20, 22, 24, 26] # note that extended grid search uses a reduced dimension grid
tol = 1e-2
valb = 5
methods = ["Simplex", "Pontos Interiores", "Híbrido"]
variables = ["num_iterations", "elapsed_time", "solution_error"] # %costfunctionerror, %error
step = 0.5
tables = {}
varcol = {}
for var in variables: # instancia lista para tabelas
    tables[var] = []
    varcol[var] = {}
    for method in methods:
        varcol[var][method] = []

for dim in dimension_grid:

    A, b, c = klee_minty(dimensions=dim, val_b=valb)

    # Simplex

    sol_simplex = simplex(A, b, c)
    simplex_var = [dim] + [sol_simplex[var] for var in variables[:-1]]
    for var in variables:
        varcol[var]["Simplex"].append(sol_simplex[var])

    simplex_var.append(0)
    optimum = sol_simplex["max_value"]

    # IP investigar porque que ta parando tao cedo se nao bateu na tolerancia

    sol_ip = interior_point(A, b, c, alpha0=step, tolerance=tol, optimum=optimum)
    solution_value_ip = sol_ip["max_value"]
    ip_value_error = (1 - (solution_value_ip/optimum))*100
    ip_var = [dim] + [sol_ip[var] for var in variables[:-1]]
    for var in variables: # substituir por compreensao de lista ein
        varcol[var]["Pontos Interiores"].append(sol_ip[var])
    ip_var.append(ip_value_error)

    # Hybrid: requires more attention, returns mutliple values for "var"

    sol_hybrid = hybrid(A, b, c, alpha0=step, tolerance=tol)
    hybrid_var = [dim] + [sol_hybrid[var]["total"] for var in variables[:-1]] + [0] # assume zero de erro
    for var in variables:
        if var=="solution_error":
            varcol[var]["Híbrido"].append(sol_hybrid[var]["simplex"])
        else:
            varcol[var]["Híbrido"].append(sol_hybrid[var]["total"])

    for v, var in enumerate(variables):
        row = []
        row.append(dim)
        row.append(simplex_var[v+1])
        row.append(ip_var[v+1])
        row.append(hybrid_var[v+1])
        tables[var].append(row)

# Pickle all data up to here! Maybe ( tables and varcol actually)

with open("tables.pickle", "wb") as handle:
    pickle.dump(tables, handle, protocol=pickle.HIGHEST_PROTOCOL)

with open("varcol.pickle", "wb") as handle:
    pickle.dump(varcol, handle, protocol=pickle.HIGHEST_PROTOCOL)