try_everything.py

import numpy as np
import os
import datetime
import game_generator
# from psro_trainer import PSRO_trainer
from utils import *
import mpmath
import copy
from itertools import product
# from meta_strategies import double_oracle,fictitious_play
from nash_solver.gambit_tools import load_pkl
from nash_solver.replicator_dynamics_solver import replicator_dynamics
from nash_solver.projected_replicator_dynamics import projected_replicator_dynamics
from scipy.stats import entropy
from MRCP.regret_analysis import extend_prob, uniform_simplex_sampling, sampling_scheme, profile_regret
from MRCP.minimum_regret_profile import minimum_regret_profile_calculator
from real_world_games.nash_clustering import nash_clustering

# meta_games = [np.random.rand(10, 10), np.random.rand(10, 10)]
# empirical_games = [[1,2,3,6], [3,5,7,9]]
# meta_games = [np.array([[1,2,3],
#                         [4,5,6],
#                         [7,8,9]]), np.array([[1,2,3],
#                         [4,5,6],
#                         [7,8,9]])]

root_path = './data/'
meta_games = load_pkl(root_path + "meta_game.pkl")
print(meta_games)
#
#
# exact_calculator = minimum_regret_profile_calculator(full_game=meta_games)
#
# empirical_games = [[12, 31, 56, 87, 111, 116, 122, 144], [4, 111, 144, 147, 158, 176, 189]]
# for _ in range(6):
#     exact_calculator.clear()
#     mrcp, regret = exact_calculator(empirical_games)
#     print(regret)

# a = mpmath.iv.exp([1, 1])
# b = mpmath.nsum(np.array(mpmath.iv.exp([1, 1])))

# aa = [mpmath.exp(100), mpmath.exp(100)]
# a = mpmath.fsum([mpmath.exp(100), mpmath.exp(100)])
# for i in aa:
#     t = i / a
#     print(t)
#     print(type(t))
#     h = np.array([np.float(t)])
#     print(h, type(h))

# a = np.array([27.586487898445515, 27.655829697611992, 24.41443100567962, 27.65316836315581, 25.765739985212146])
# b = np.array([0.66666667, 0.11111111, 0.11111111, 0.11111111, 0.        ])
# print(np.sum(a * b))
# print(np.max(a) - np.sum(a * b))