Merge pull request #773 from Axelrod-Python/ann-mojones

Merge evolved artificial neural network strategy

Author: meatballs

axelrod/strategies/_strategies.py

@@ -2,6 +2,7 @@
 
 from .alternator import Alternator
 from .adaptive import Adaptive
+from .ann import EvolvedANN
 from .apavlov import APavlov2006, APavlov2011
 from .appeaser import Appeaser
 from .averagecopier import AverageCopier, NiceAverageCopier
@@ -105,6 +106,8 @@
     EasyGo,
     Eatherley,
     EventualCycleHunter,
+    EvolvedANN,
+    EvolvedLookerUp,
     Feld,
     FirmButFair,
     FoolMeForever,
@@ -150,7 +153,6 @@
     LimitedRetaliate,
     LimitedRetaliate2,
     LimitedRetaliate3,
-    EvolvedLookerUp,
     MathConstantHunter,
     NaiveProber,
     MindBender,

axelrod/strategies/ann.py

@@ -0,0 +1,230 @@
+# Source: https://gist.github.com/mojones/550b32c46a8169bb3cd89d917b73111a#file-ann-strategy-test-L60
+# Original Author: Martin Jones, @mojones
+
+from axelrod import Actions, Player, init_args
+
+C, D = Actions.C, Actions.D
+
+
+def split_weights(weights, input_values, hidden_layer_size):
+    """Splits the input vector into the the NN bias weights and layer
+    parameters."""
+    number_of_input_to_hidden_weights = input_values * hidden_layer_size
+    number_of_hidden_to_output_weights = hidden_layer_size
+
+    input2hidden = []
+    for i in range(0, number_of_input_to_hidden_weights, input_values):
+        input2hidden.append(weights[i:i + input_values])
+
+    start = number_of_input_to_hidden_weights
+    end = number_of_input_to_hidden_weights + number_of_hidden_to_output_weights
+
+    hidden2output = weights[start: end]
+    bias = weights[end:]
+
+    return (input2hidden, hidden2output, bias)
+
+
+class ANN(Player):
+    """A single layer neural network based strategy."""
+    name = 'ANN'
+    classifier = {
+        'memory_depth': float('inf'),
+        'stochastic': False,
+        'inspects_source': False,
+        'makes_use_of': set(),
+        'manipulates_source': False,
+        'manipulates_state': False,
+        'long_run_time': False
+    }
+
+    def activate(self, inputs):
+        """Compute the output of the neural network."""
+        # Calculate values of hidden nodes
+        hidden_values = []
+        for i in range(self.hidden_layer_size):
+            hidden_node_value = 0
+            bias_weight = self.bias_weights[i]
+            hidden_node_value += bias_weight
+            for j in range(self.input_values):
+                weight = self.input_to_hidden_layer_weights[i][j]
+                hidden_node_value += inputs[j] * weight
+
+            # ReLU activation function
+            hidden_node_value = max(hidden_node_value, 0)
+
+            hidden_values.append(hidden_node_value)
+
+        # Calculate output value
+        output_value = 0
+        for i in range(self.hidden_layer_size):
+            output_value += hidden_values[i] * \
+                            self.hidden_to_output_layer_weights[i]
+
+        return output_value
+
+    @init_args
+    def __init__(
+        self,
+        input_to_hidden_layer_weights=[],
+        hidden_to_output_layer_weights=[],
+        bias_weights=[]
+    ):
+
+        Player.__init__(self)
+        self.input_to_hidden_layer_weights = input_to_hidden_layer_weights
+        self.hidden_to_output_layer_weights = hidden_to_output_layer_weights
+        self.bias_weights = bias_weights
+
+        self.input_values = len(input_to_hidden_layer_weights[0])
+        self.hidden_layer_size = len(hidden_to_output_layer_weights)
+
+    def strategy(self, opponent):
+        # Compute features for Neural Network
+        # These are True/False 0/1
+        if len(opponent.history) == 0:
+            opponent_first_c = 0
+            opponent_first_d = 0
+            opponent_second_c = 0
+            opponent_second_d = 0
+            my_previous_c = 0
+            my_previous_d = 0
+            my_previous2_c = 0
+            my_previous2_d = 0
+            opponent_previous_c = 0
+            opponent_previous_d = 0
+            opponent_previous2_c = 0
+            opponent_previous2_d = 0
+
+        elif len(opponent.history) == 1:
+            opponent_first_c = 1 if opponent.history[0] == C else 0
+            opponent_first_d = 1 if opponent.history[0] == D else 0
+            opponent_second_c = 0
+            opponent_second_d = 0
+            my_previous_c = 1 if self.history[-1] == C else 0
+            my_previous_d = 0 if self.history[-1] == D else 0
+            my_previous2_c = 0
+            my_previous2_d = 0
+            opponent_previous_c = 1 if opponent.history[-1] == C else 0
+            opponent_previous_d = 1 if opponent.history[-1] == D else 0
+            opponent_previous2_c = 0
+            opponent_previous2_d = 0
+
+        else:
+            opponent_first_c = 1 if opponent.history[0] == C else 0
+            opponent_first_d = 1 if opponent.history[0] == D else 0
+            opponent_second_c = 1 if opponent.history[1] == C else 0
+            opponent_second_d = 1 if opponent.history[1] == D else 0
+            my_previous_c = 1 if self.history[-1] == C else 0
+            my_previous_d = 0 if self.history[-1] == D else 0
+            my_previous2_c = 1 if self.history[-2] == C else 0
+            my_previous2_d = 1 if self.history[-2] == D else 0
+            opponent_previous_c = 1 if opponent.history[-1] == C else 0
+            opponent_previous_d = 1 if opponent.history[-1] == D else 0
+            opponent_previous2_c = 1 if opponent.history[-2] == C else 0
+            opponent_previous2_d = 1 if opponent.history[-2] == D else 0
+
+        # Remaining Features
+        turns_remaining = self.match_attributes['length'] - len(self.history)
+        total_opponent_c = opponent.history.count(C)
+        total_opponent_d = opponent.history.count(D)
+        total_self_c = self.history.count(C)
+        total_self_d = self.history.count(D)
+
+        output = self.activate([
+            opponent_first_c,
+            opponent_first_d,
+            opponent_second_c,
+            opponent_second_d,
+            my_previous_c,
+            my_previous_d,
+            my_previous2_c,
+            my_previous2_d,
+            opponent_previous_c,
+            opponent_previous_d,
+            opponent_previous2_c,
+            opponent_previous2_d,
+            total_opponent_c,
+            total_opponent_d,
+            total_self_c,
+            total_self_d,
+            turns_remaining
+        ])
+        if output > 0:
+            return C
+        else:
+            return D
+
+
+class EvolvedANN(ANN):
+    """
+    A strategy based on a pre-trained neural network.
+
+    Names:
+
+     - EvolvedANN: : Original name by Martin Jones.
+    """
+
+    name = "EvolvedANN"
+
+    @init_args
+    def __init__(self):
+        self.classifier['makes_use_of'] = set(['length'])
+        input_values = 17
+        hidden_layer_size = 10
+
+        weights = [0.19789658035994948, -5575.476236516673, 0.1028948855131803, 0.7421752484224489,
+                       -16.286246197005298, 11708.007255945553, 0.01400184611448853, -33.39126355009626,
+                       -12.755203414662356, -32.92388754142929, 197.3517717772447, 108262.87038790248,
+                       -0.1084768512582505, 85.20738888799768, 723.9537664890132, -2.59453614458083,
+                       0.5599936275978272, 7.89217571665664, -48014.821440080384, -1.364025168184463,
+                       -1.062138244222801, 11153713.883580556, -59.58314524751318, 51278.916519524784,
+                       3196.528224457722, -4635.771421694692, -129.93354968926164, -0.7927383528469051,
+                       98.47779304649353, -81.19056440190543, 29.53082483602472, -48.16562780387682,
+                       49.40755170297665, 288.3295763937912, -68.38780651250116, -167.64039570334904,
+                       -0.1576073061122998, 160.6846658333963, 34.55451693336857, -0.08213997499783675,
+                       -4.802560347075611, -1.4042000430302104, -0.9832145174590058, 0.008705149387813573,
+                        14.041842191255089, 0.05395665905821821, -0.13856885306885558, 5.311455433711278,
+                       -5.835498171845142, 0.00010294700612334848, 26.42528200366623, 33.690839945794785,
+                       7.931017950666591, -0.00037662122944226125, 59.295075951374606, -0.15888507169191035,
+                       3.670332254391659, 789.6230735057893, -0.7367125124436135, -198.44119280589902,
+                       537.9939493545736, -287.54344903637207, 1759.5455359353778, -18.48997020629342,
+                       -8426184.81603275, -82.36805426730088, 1144.1032034358543, 15635.402592538396,
+                       3095.643889329041, 2332.107673930774, -0.5601648316602144, 101.98300711150003,
+                       -7387.135294747112, -4241.004613717573, 3.06175607282536e-05, -35122.894421260884,
+                       -38591.45572476855, -0.16081285130591272, -29608.73087879185, 122.47563639056185,
+                       6.381946054740736, -0.8978628581801188, 17658.47647781355, -0.011719257684286711,
+                       0.10734295104044986, -378.35448968529494, 225.06912279045062, -351.12326495980847,
+                       -1.927322672845826, 0.0014584395475859544, -8.629826916169318, 22.43281153854352,
+                       87.10895591188721, -0.22253937914423294, -233.06796470563208, -620.4917481128365,
+                       -1.8253699204909606,-0.0030318160426064467, -77.25818476745101, -2057.311059352977,
+                       3.764204074005541, -47.47629147374066, 233.16096124330778, -160721.96744375565,
+                       -278292.9688140893, -2.093640525920404, -142886.66171202937, 53.64449245132945,
+                       12.5162147724691, -207.75462390139955, 132167.659160016, 21.197418541051732,
+                       83979.45623573882, -49.47558832987566, 0.05242625398046057, -842.1484416713075,
+                       -0.1581049310461208, 2359.2124343564096, 1170.5147830681053, -847999.9145843558,
+                       -0.8053911061885284, -5363.722820739466, 171.58433274294117, -724.7468082647013,
+                       2500359.853524033, 1595.3955511798079, -4.254009123616706, -171.12968391407912,
+                       -32.30624102753424, -558.412338112568, -234.29754199019308, -18768.34057250429,
+                       8338.792126484348, -0.18593140210730602, -7.758804964874875, 0.39736677884665267,
+                       547.0567585452197, 1.1969366369973133, 0.4861465741177498, -51.19319208716985,
+                       12.775051406025534, -0.09185362260212569, 22.08417300332754, -5090.013231748707,
+                       -0.814394991797045, 1.1534025840023847, 8.390439959276764, -0.02227253403481858,
+                       0.14162040507921927, -0.011508263843203926, 0.22372493104861083, 0.7754713610627112,
+                       0.1044033140236981, -4.377055307648915, -41.898221495326574, -18656.755601828827,
+                       -134.56719406539244, -2405.8148785743474, 16864.049985157206, -0.5124682025216784,
+                       14521.069005125159, -10.740782200739309, 18756.807715014013, -1723.9353962656946,
+                       87029.99828299093, 5.7383786020894195e-05, 4762.960401619296, 0.7331769713238158,
+                       -308.5673034493341, 85.29725765515369, 0.4268843538235295, -0.17788805472511407,
+                       -1.1727033611646802, 7578.6822604990175, 0.5124673187864222, 0.1595627909684813,
+                       -145.93742731401096, -2954.234440189563, 0.009672881359732015, 106.4646644917487,
+                       -0.050606976105730346, 2.3904047264403596, -4.987645640997455, -43.22984692765006,
+                       -36.177108409134966, -0.3812547430698569, -2959.4921368963633, -1.8635802741029985,
+                       0.020513128847167047, -0.9179124323385958]
+
+        (i2h, h2o, bias) = split_weights(
+            weights,
+            input_values,
+            hidden_layer_size
+        )
+        ANN.__init__(self, i2h, h2o, bias)

axelrod/tests/unit/test_ann.py

@@ -0,0 +1,43 @@
+"""Test for the Adaptive strategy."""
+
+import axelrod
+
+from .test_player import TestHeadsUp, TestPlayer
+
+C, D = axelrod.Actions.C, axelrod.Actions.D
+
+
+class TestEvolvedANN(TestPlayer):
+
+    name = "EvolvedANN"
+    player = axelrod.EvolvedANN
+    expected_classifier = {
+        'memory_depth': float('inf'),
+        'stochastic': False,
+        'makes_use_of': set(["length"]),
+        'long_run_time': False,
+        'inspects_source': False,
+        'manipulates_source': False,
+        'manipulates_state': False
+    }
+
+    def test_strategy(self):
+        # Test initial play sequence
+        self.first_play_test(C)
+
+
+class TestEvolvedANNvsCooperator(TestHeadsUp):
+    def test_rounds(self):
+        self.versus_test(axelrod.EvolvedANN(), axelrod.Cooperator(),
+                         [C, D, D, C, D], [C] * 5)
+
+
+class TestEvolvedANNvsDefector(TestHeadsUp):
+    def test_rounds(self):
+        self.versus_test(axelrod.EvolvedANN(), axelrod.Defector(),
+                         [C, D, D, D, D], [D] * 5)
+
+class TestEvolvedANNvsTFT(TestHeadsUp):
+    def test_rounds(self):
+        self.versus_test(axelrod.EvolvedANN(), axelrod.TitForTat(),
+                         [C, D, D, C, C], [C, C, D, D, C] * 5)

docs/tutorials/advanced/classification_of_strategies.rst

@@ -81,7 +81,7 @@ length of each match of the tournament::
     ... }
     >>> strategies = axl.filtered_strategies(filterset)
     >>> len(strategies)
-    5
+    6
 
 Note that in the filterset dictionary, the value for the 'makes_use_of' key
 must be a list. Here is how we might identify the number of strategies that use