def setUp(self):
'''
Initializes the necessary resources for the tests.
Works only if IS_SIGMA_RANDOM == 0 in global_values.py
The expected sequence of generated random numbers with seed 10 is:
random.seed(10)
random.uniform(0,20) ------- the map is 20 in each sides
random.uniform(0,5) --- for the start of the cum_knonwledge
the values of the agent 0 are:
map_market = 11.4280518938
map_knowledge = 8.5777810935
cum_knowledge = 2.89045650567
the values of the agent 1 are:
map_market = 4.12196464279
map_knowledge = 16.2664250271
cum_knowledge = 4.11794436267
the values of the agent 2 are:
map_market = 13.069450678
map_knowledge = 3.20459113038
cum_knowledge = 2.6033467982
the values of the agent 3 are:
map_market = 6.55545623244
map_knowledge = 4.99993353373
cum_knowledge = 4.76408454573
the values of the agent 4 are:
map_market = 19.9311398508
map_knowledge = 0.891127649009
cum_knowledge = 4.30080518643
the values of the agent 5 are:
map_market = 12.0638122194
map_knowledge = 7.63211971838
cum_knowledge = 1.41809108953
'''
random.seed(10)
self.cyclic_data = CyclicData()
self.network = Network(6, self.cyclic_data)
if not os.path.exists(OUTPUT_DIR):
os.makedirs(OUTPUT_DIR)
from main.network import Network net = Network([3, 2, 3]) traininig_data = list(zip(range(1, 10), range(10, 1, -1))) epochs = 5 mini_batch_size = 2 eta = 0.01 test_data = [] net.SGD(traininig_data, epochs, mini_batch_size, eta, test_data)
class TestNetwork(unittest.TestCase):
def setUp(self):
'''
Initializes the necessary resources for the tests.
Works only if IS_SIGMA_RANDOM == 0 in global_values.py
The expected sequence of generated random numbers with seed 10 is:
random.seed(10)
random.uniform(0,20) ------- the map is 20 in each sides
random.uniform(0,5) --- for the start of the cum_knonwledge
the values of the agent 0 are:
map_market = 11.4280518938
map_knowledge = 8.5777810935
cum_knowledge = 2.89045650567
the values of the agent 1 are:
map_market = 4.12196464279
map_knowledge = 16.2664250271
cum_knowledge = 4.11794436267
the values of the agent 2 are:
map_market = 13.069450678
map_knowledge = 3.20459113038
cum_knowledge = 2.6033467982
the values of the agent 3 are:
map_market = 6.55545623244
map_knowledge = 4.99993353373
cum_knowledge = 4.76408454573
the values of the agent 4 are:
map_market = 19.9311398508
map_knowledge = 0.891127649009
cum_knowledge = 4.30080518643
the values of the agent 5 are:
map_market = 12.0638122194
map_knowledge = 7.63211971838
cum_knowledge = 1.41809108953
'''
random.seed(10)
self.cyclic_data = CyclicData()
self.network = Network(6, self.cyclic_data)
if not os.path.exists(OUTPUT_DIR):
os.makedirs(OUTPUT_DIR)
# for i in range(6):
# print 'Agent = ', i
# print 'market = ', self.network.agents[i].map_market
# print 'knowledge = ', self.network.agents[i].map_knowledge
# print 'cum_know = ', self.network.agents[i].cum_knowledge
# print '----------------'
def tearDown(self):
self.network = None
self.cyclic_data = None
if os.path.exists(OUTPUT_DIR):
os.removedirs(OUTPUT_DIR)
def test_constructor(self):
'''
Tests the constructor of the Network class.
'''
self.assertEqual(0, self.network.cyclic_data.cycle_number)
self.assertEqual(6, len(self.network.agents))
self.assertNotEqual(None, self.network.sigma)
self.assertEqual(20.094728488233791, self.network.total_cum_knowledge)
self.assertEqual(3.3491214147056318, self.network.average_agent_cum_knowledge)
self.assertEqual(0, len(self.network.expected_learning_matrix))
self.assertEqual(0, len(self.network.expected_learning_matrix_with_loss))
# the values of the agent 0 are:
# map_market = 11.4280518938
# map_knowledge = 8.5777810935
# cum_knowledge = 2.89045650567
self.assertAlmostEqual(11.4280518938, \
self.network.agents[0].map_market)
self.assertAlmostEqual(8.5777810935, \
self.network.agents[0].map_knowledge)
self.assertAlmostEqual(2.8904565056723519, \
self.network.agents[0].cum_knowledge)
# the values of the agent 1 are:
# map_market = 4.12196464279
# map_knowledge = 16.2664250271
# cum_knowledge = 4.11794436267
self.assertAlmostEqual(4.12196464279, \
self.network.agents[1].map_market)
self.assertAlmostEqual(16.2664250271, \
self.network.agents[1].map_knowledge)
self.assertAlmostEqual(4.11794436267, \
self.network.agents[1].cum_knowledge)
# the values of the agent 2 are:
# map_market = 13.069450678
# map_knowledge = 3.20459113038
# cum_knowledge = 2.6033467982
self.assertAlmostEqual(13.069450678, \
self.network.agents[2].map_market)
self.assertAlmostEqual(3.20459113038, \
self.network.agents[2].map_knowledge)
self.assertAlmostEqual(2.6033467982, \
self.network.agents[2].cum_knowledge)
# the values of the agent 3 are:
# map_market = 6.55545623244
# map_knowledge = 4.99993353373
# cum_knowledge = 4.76408454573
self.assertAlmostEqual(6.55545623244, \
self.network.agents[3].map_market)
self.assertAlmostEqual(4.99993353373, \
self.network.agents[3].map_knowledge)
self.assertAlmostEqual(4.76408454573, \
self.network.agents[3].cum_knowledge)
# the values of the agent 4 are:
# map_market = 19.9311398508
# map_knowledge = 0.891127649009
# cum_knowledge = 4.30080518643
self.assertAlmostEqual(19.9311398508, \
self.network.agents[4].map_market)
self.assertAlmostEqual(0.891127649009, \
self.network.agents[4].map_knowledge)
self.assertAlmostEqual(4.30080518643, \
self.network.agents[4].cum_knowledge)
# the values of the agent 5 are:
# map_market = 12.0638122194
# map_knowledge = 7.63211971838
# cum_knowledge = 1.41809108953
self.assertAlmostEqual(12.0638122194, \
self.network.agents[5].map_market)
self.assertAlmostEqual(7.63211971838, \
self.network.agents[5].map_knowledge)
self.assertAlmostEqual(1.41809108953, \
self.network.agents[5].cum_knowledge)
self.assertAlmostEqual(20.094728488233791, self.network.total_cum_knowledge)
def test_reset(self):
'''
Tests the reset method of the Network class.
'''
self.network.cyclic_data.set_cycle(5)
self.network.agents[1].exit()
self.network.reset()
self.assertEqual(0, self.network.average_agent_cum_knowledge)
self.assertEqual(0, len(self.network.expected_learning_matrix))
self.assertEqual(0, len(self.network.expected_learning_matrix_with_loss))
self.assertEqual(5, self.network.cyclic_data.cycle_number)
self.assertEqual(6, len(self.network.agents))
self.assertNotEqual(None, self.network.sigma)
self.assertEqual(20.094728488233791, self.network.total_cum_knowledge)
self.assertEqual(0.0, self.network.agents[0].cycle_realized_learning)
self.assertEqual(None, self.network.agents[0].alliance)
self.assertEqual(0.0, self.network.agents[0].next_map_market)
self.assertEqual(0.0, self.network.agents[0].next_map_knowledge)
self.assertEqual(0.0, self.network.agents[5].cycle_realized_learning)
self.assertEqual(None, self.network.agents[5].alliance)
self.assertEqual(0.0, self.network.agents[5].next_map_market)
self.assertEqual(0.0, self.network.agents[5].next_map_knowledge)
self.assertAlmostEqual(4.12196464279, \
self.network.agents[1].map_market)
self.assertAlmostEqual(16.2664250271, \
self.network.agents[1].map_knowledge)
def test_create_expected_learning_matrix(self):
'''
Tests the create_evaluation_matrices method of the Network class.
'''
# agent 2 is exit so every expected learning calculations
# related to agent 2 should be 0
self.network.agents[2].exit()
self.network.create_evaluation_matrices()
# The expected learning of agent2 from any other agent must be
# 0 since agent2 is not active.
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[0][2])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[1][2])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[3][2])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[4][2])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[5][2])
# The expected learning of agent0 from agent0 must be 0.
self.assertEqual(0, self.network.expected_learning_matrix[0][0])
self.assertEqual(0, self.network.expected_learning_matrix[1][1])
self.assertEqual(0, self.network.expected_learning_matrix[2][2])
self.assertEqual(0, self.network.expected_learning_matrix[3][3])
self.assertEqual(0, self.network.expected_learning_matrix[4][4])
self.assertEqual(0, self.network.expected_learning_matrix[5][5])
# Test expected_learning for agent0
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[0][1])
self.assertAlmostEqual(14.1451284754856, \
self.network.expected_learning_matrix[0][3])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[0][4])
self.assertAlmostEqual(3.08468113368173, \
self.network.expected_learning_matrix[0][5])
# Test expected_learning for agent 1
self.assertAlmostEqual(2.52758171621341, \
self.network.expected_learning_matrix[1][0])
self.assertAlmostEqual(8.5659563015936, \
self.network.expected_learning_matrix[1][3])
self.assertAlmostEqual(16.5230281780298, \
self.network.expected_learning_matrix[1][4])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[1][5])
# Test expected_learning for agent 2 should be zero because
# agent 2 is exit
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[2][0])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[2][1])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[2][3])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[2][4])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[2][5])
# Test expected_learning for agent 3
self.assertAlmostEqual(17.7139840641858, \
self.network.expected_learning_matrix[3][0])
self.assertAlmostEqual(0.956444371563587, \
self.network.expected_learning_matrix[3][1])
self.assertAlmostEqual(16.1349628236964, \
self.network.expected_learning_matrix[3][4])
self.assertAlmostEqual(16.2426173253092, \
self.network.expected_learning_matrix[3][5])
# Test expected_learning for agent 4
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[4][0])
self.assertAlmostEqual(16.7010192277332, \
self.network.expected_learning_matrix[4][1])
self.assertAlmostEqual(11.8150724425493, \
self.network.expected_learning_matrix[4][3])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[4][5])
# Test expected_learning for agent 5
self.assertAlmostEqual(0.905351406607448, \
self.network.expected_learning_matrix[5][0])
self.assertAlmostEqual(0, \
self.network.expected_learning_matrix[5][1])
self.assertAlmostEqual(5.44885016295865, \
self.network.expected_learning_matrix[5][3])
self.assertAlmostEqual(7.55144184619663, \
self.network.expected_learning_matrix[5][4])
####
# Tests the expected learning after loss of agent0 from agent5 at radius R.
# self.assertEqual(1.4688863524797151, self.network.expected_learning_matrix_with_loss[0][5])
def test_calculate_learning_after_loss(self):
'''
Tests the calculate_learning_after_loss method of the Network class.
'''
# num_agent = len(self.network.agents)
# self.network.create_evaluation_matrices()
# self.network.expected_learning_matrix_with_loss = Calculator.init_matrix(num_agent, 0)
# self.assertEqual(1.4688863524797151, self.network.calculate_learning_after_loss(5.7299199550382705, 4.2652490123731477, 0, 5, 1.4688863524797151))
# Test with R = 2
#self.assertAlmostEqual(1.4688863524797151, self.network.calculate_learning_after_loss(5.7299199550382705, 4.2652490123731477, 0, 5, 1.4688863524797151))
def test_evaluate_alliances(self):
'''
Tests the evaluate_alliances method of the Network class.
'''
pass
def test_try_alliance(self):
'''
Tests the try_alliance method of the Network class.
'''
pass
def test_can_make_alliance(self):
'''
Tests the can_make_alliance method of the Network class.
'''
pass
def test_calculate_realized_learning(self):
'''
Tests the calculate_realized_learning method of the Network class.
'''
pass
def test_manage_exit(self):
'''
Tests the manage_exit method of the Network class.
'''
def test_plot_map(self):
'''
Tests the plot_map method of the Network class.
'''
self.network.plot_map(0, 0)
def test_plot_map_exception(self):
'''
Tests the plot_map method of the Network class for invalid input.
'''
self.assertRaises(InputError, Network.plot_map, self.network, -1, 0)
self.assertRaises(InputError, Network.plot_map, self.network, 1, -2)