def train_the_net(self):
start_time = time.time()
# Reading the inputs from a file
logging.info("--- Training patterns ---")
logging.info("-----------------------------------------------------------")
logging.info(" Pattern\t\tTarget")
for line in open('train.txt', 'r'):
self.training_patterns.append(np.matrix(line.split('|')[0]))
self.training_targets.append(np.matrix(line.split('|')[1]))
logging.info("%s", line.replace('\n', '').replace('|', '\t\t'))
# Creating & training the network
input_length = len(self.training_patterns[0])
target_length = len(self.training_targets[0])
self.net = Net(input_length, target_length)
self.net.learn(self.training_patterns, self.training_targets, self.desired_error)
# If alpha is not set, it would be calculated by the correlation matrix
# You can also add a weights matrix and a threshold value to the network and add them to the learn() function. Eg
# net.learn(self.training_patterns, self.training_targets, self.desired_error, alpha=0.4, weights=w, threshold=b)
# Now the ADALINE is trained and we can get the results and save them in a file
f = open('weights.txt', 'w')
f.write('W = ' + self.net.weights.__str__() + '\n')
f.write('b = ' + self.net.threshold.__str__())
f.close()
# Showing weights and threshold in GUi
self.label.set(
'alpha = ' + self.net.alpha.__str__() + '\nW = ' + self.net.weights.__str__() + '\nb = ' + self.net.threshold.__str__())
# Plotting the inputs and targets
logging.info("--- Training finished in %s seconds ---", (time.time() - start_time))
logging.info("-----------------------------------------------------------")
Plotter().plot3d(self.training_patterns, self.training_targets, 'Training patterns', weights=self.net.weights,
threshold=self.net.threshold)
class Main:
def __init__(self):
logging.basicConfig(filename='adaline.log', level=logging.DEBUG)
logging.getLogger().addHandler(logging.StreamHandler())
self.training_patterns = []
self.training_targets = []
self.desired_error = 1e-10
if desired_error is not None:
self.desired_error = desired_error
self.alpha = alpha
logging.info("-----------------------------------------------------------")
logging.info(" A D A L I N E E X A M P L E")
logging.info("-----------------------------------------------------------")
# Creating GUI
top = Tkinter.Tk()
top.wm_title("ADALINE Example")
Tkinter.Button(top, text="Entrenar red", command=self.train_the_net).pack()
Tkinter.Button(top, text="Clasificar patrones", command=self.classify_patterns).pack()
self.label = Tkinter.StringVar()
Tkinter.Label(top, textvariable=self.label).pack()
top.mainloop()
def train_the_net(self):
start_time = time.time()
# Reading the inputs from a file
logging.info("--- Training patterns ---")
logging.info("-----------------------------------------------------------")
logging.info(" Pattern\t\tTarget")
for line in open('train.txt', 'r'):
self.training_patterns.append(np.matrix(line.split('|')[0]))
self.training_targets.append(np.matrix(line.split('|')[1]))
logging.info("%s", line.replace('\n', '').replace('|', '\t\t'))
# Creating & training the network
input_length = len(self.training_patterns[0])
target_length = len(self.training_targets[0])
self.net = Net(input_length, target_length)
self.net.learn(self.training_patterns, self.training_targets, self.desired_error)
# If alpha is not set, it would be calculated by the correlation matrix
# You can also add a weights matrix and a threshold value to the network and add them to the learn() function. Eg
# net.learn(self.training_patterns, self.training_targets, self.desired_error, alpha=0.4, weights=w, threshold=b)
# Now the ADALINE is trained and we can get the results and save them in a file
f = open('weights.txt', 'w')
f.write('W = ' + self.net.weights.__str__() + '\n')
f.write('b = ' + self.net.threshold.__str__())
f.close()
# Showing weights and threshold in GUi
self.label.set(
'alpha = ' + self.net.alpha.__str__() + '\nW = ' + self.net.weights.__str__() + '\nb = ' + self.net.threshold.__str__())
# Plotting the inputs and targets
logging.info("--- Training finished in %s seconds ---", (time.time() - start_time))
logging.info("-----------------------------------------------------------")
Plotter().plot3d(self.training_patterns, self.training_targets, 'Training patterns', weights=self.net.weights,
threshold=self.net.threshold)
def classify_patterns(self):
start_time = time.time()
# Once the network is trained, classify some patterns
self.patterns = []
self.targets = []
for line in open('inputs.txt', 'r'):
self.patterns.append(np.matrix(line.split('|')[0]))
self.targets = self.net.classify(self.patterns)
# Now we can get the results
logging.info("-----------------------------------------------------------")
logging.info("--- Pattern classification results ---")
logging.info("-----------------------------------------------------------")
logging.info(" Pattern \t\t Target")
f = open('results.txt', 'w')
tar = [[int(round(q,0)) for q in p] for p in self.targets]
for p, t in zip(self.patterns, tar):
f.write(p.__str__().replace('[', '').replace(']', '').replace('\n', ';') + "|" + t.__str__().replace('[','').replace(']','').replace(',',';') + "\n")
logging.info(p.__str__().replace('[', '').replace(']', '').replace('\n', ';') + "\t" + t.__str__().replace('[','').replace(']','').replace(',',';'))
f.close()
logging.info("--- Classifying finished in %s seconds ---", (time.time() - start_time))
logging.info("-----------------------------------------------------------")
# Plotting the inputs and targets
Plotter().plot3d(self.patterns, self.targets, 'Pattern classification', weights=self.net.weights,
threshold=self.net.threshold)
