def __init__(self, games_count=50, nodes=[100], preprocess=False):
self.is_preprocess = preprocess
# Load training data
data_2048 = np.loadtxt('data' + system_divider + 'log-2048-demo.txt', dtype=float, usecols=range(17))
# Get the labels
raw_labels_2048 = data_2048[:, 16]
labels_2048 = [[0 for y in range(4)] for x in range(len(raw_labels_2048))]
for i in range(len(raw_labels_2048)):
labels_2048[i][int(raw_labels_2048[i])] = 1.0
labels_2048 = np.array(labels_2048)
# Get the states
states_2048 = np.delete(data_2048, np.s_[-1:], 1)
# Preprocess training data
if self.is_preprocess:
states_2048 = self.preprocess(states_2048)
# Wrap the data
data = [states_2048, labels_2048, states_2048,
labels_2048]
# Get the imput size
input_size = np.array(states_2048).shape[1]
if nodes != [100]:
print('Layers ' + str(nodes))
# Initialize neural network
self.neural_net = ANN(nodes=[24] + nodes + [4], data=data)
# Train
self.train()
# Call super
super(Neural, self).__init__(games_count=games_count)
from network import ANN from compare import * from train import Trainer import numpy as np x = np.array([[4, 5.5], [4.5, 1], [9, 2.5], [6,2], [10, 10]], dtype = float) y = np.array([[70], [89], [85], [75], [10]], dtype = float) N = ANN(2, 5, 1, 0.000001) x = x / np.amax(x) y = y / np.amax(y) print "y -> net(y)" print y, "\n------------\n", N.forward(x) ag = analyticalGrad(N, x, y) ng = numericalGrad (N, x, y) print "Analytical grad: ", sum(ag), ", ", ag.shape print "Numerical grad: ", sum(ng), ", ", ng.shape T = Trainer(N) T.train(x,y) ag = analyticalGrad(N, x, y) ng = numericalGrad (N, x, y) print "Analytical grad: ", sum(ag), ", ", ag.shape print "Numerical grad: ", sum(ng), ", ", ng.shape
class Neural(Player):
neural_net = None
def __init__(self, games_count=50, nodes=[100], preprocess=False):
self.is_preprocess = preprocess
# Load training data
data_2048 = np.loadtxt('data' + system_divider + 'log-2048-demo.txt', dtype=float, usecols=range(17))
# Get the labels
raw_labels_2048 = data_2048[:, 16]
labels_2048 = [[0 for y in range(4)] for x in range(len(raw_labels_2048))]
for i in range(len(raw_labels_2048)):
labels_2048[i][int(raw_labels_2048[i])] = 1.0
labels_2048 = np.array(labels_2048)
# Get the states
states_2048 = np.delete(data_2048, np.s_[-1:], 1)
# Preprocess training data
if self.is_preprocess:
states_2048 = self.preprocess(states_2048)
# Wrap the data
data = [states_2048, labels_2048, states_2048,
labels_2048]
# Get the imput size
input_size = np.array(states_2048).shape[1]
if nodes != [100]:
print('Layers ' + str(nodes))
# Initialize neural network
self.neural_net = ANN(nodes=[24] + nodes + [4], data=data)
# Train
self.train()
# Call super
super(Neural, self).__init__(games_count=games_count)
def preprocess(self, data):
shape = np.array(data).shape
# If list of flat lists
if shape[0] > shape[1]:
copy = []
for i in range(shape[0]):
copy.append(
Process.multiple_methods(argument=data[i], methods=[Process.logarithm, Process.moveable_lines]))
return copy
# If one single
else:
return Process.multiple_methods(argument=data, methods=[Process.logarithm, Process.moveable_lines])
def train(self, batch=20, verbose_level=0, epochs=10):
"""
Method for training the network
"""
print('Training ...')
self.neural_net.train(batch=batch, verbose_level=verbose_level, epochs=epochs)
def do_move(self, game):
prediction = None
# Should preprocess ?
if self.is_preprocess:
# Preprocess to match training
n = self.preprocess(game.grid)
# Do move
prediction = self.neural_net.blind_test([n])[0]
else:
flat = [np.array(game.grid).flatten()]
prediction = self.neural_net.blind_test(flat)[0]
# print(prediction)
for pred in prediction:
if game.move(self.directions[pred]):
return True
# print('Game over')
# print(game.grid)
# No lleagal moves, game over
return False
training = np.array([ ['10000000', '10000000'], ['01000000', '01000000'], ['00100000', '00100000'], ['00010000', '00010000'], ['00001000', '00001000'], ['00000100', '00000100'], ['00000010', '00000010'], ['00000001', '00000001']]) #training = np.array([ ['10000000', '10000000']]) nn = ANN(num_features, num_labels, num_hidden) # Build training features training_features = [] for i in range (len(training)): training_features.append( np.asarray( list(training[i][0]), dtype = int )) training_features = np.asarray(training_features) # Build training labels training_labels = [] for i in range (len(training)): training_labels.append( list(np.asarray( list(training[i][1]), dtype = int ))) training_labels = np.asarray(training_labels)
