Data Science Asked by Szafran on May 5, 2021
I’ve done perceptron with one output already, but now i want to try doing it with at least two outputs. I can’t find any example in google. Is it possible? Can you give me a minimal example? I want to create neural network which is predicting binary numbers. I don’t know how to add more outputs. It will be the best option if i can create 8 output neurons.
Thats what i’ve written:
def sigmoid(x):
return 1.0 / (1 + exp(-x))
def sigmoid_deriv(x):
return sigmoid(x) * (1 - sigmoid(x))
training_data = [
(array([0,0,1]), [0, 0]),
(array([1,0,1]), [1, 0]),
(array([0,1,1]), [0, 1]),
(array([1,1,1]), [1, 1]),
]
def binary_to_int(binary_number_array: list):
#print(binary)
decimal = 0
i = 0
#print(binary_number_array)
for number in binary_number_array:
if number == 1:
decimal += pow(2, i)
i += 1
#print(decimal)
return decimal
errors = []
eta = 0.6
input_layer = 3
hidden_layer = 2
output_layer = 2
epoch_count = 500
weights_input_to_hidden = np.random.uniform(-1, 1, (input_layer, hidden_layer))
weights_hidden_to_output = np.random.uniform(-1, 1, hidden_layer)
preActivation_H = np.zeros(hidden_layer)
postActivation_H = np.zeros(hidden_layer)
training_count = len(training_data[:])
#print(training_count)
for epoch in range(epoch_count):
for sample in range(training_count):
data, expected = training_data[sample]
for node in range(hidden_layer):
#data, expected = training_data[sample]
#print(data, weights_input_to_hidden[:, node])
preActivation_H[node] = np.dot(data, weights_input_to_hidden[:, node])
postActivation_H[node] = sigmoid(preActivation_H[node])
#print(data, weights_input_to_hidden[:, node])
#print(preActivation_H[node], postActivation_H[node])
preActivation_O = np.dot(postActivation_H, weights_hidden_to_output)
postActivation_O = sigmoid(preActivation_O)
decimal_expected = binary_to_int(expected)
error = postActivation_O - decimal_expected
errors.append(error)
#print(error)
# back propagation
for hidden_node in range(hidden_layer):
S_error = error * sigmoid_deriv(preActivation_O)
#print(S_error)
gradient_hidden_to_output = S_error * postActivation_H[hidden_node]
#print(gradient_hidden_to_output)
for input_node in range(input_layer):
# tu moze byc zle
input_value = data[input_node]
print(input_value)
gradient_input_to_hidden = S_error * weights_hidden_to_output[hidden_node] *
sigmoid_deriv(preActivation_H[hidden_node]) * input_value
weights_input_to_hidden[input_node, hidden_node] -= eta * gradient_input_to_hidden
weights_hidden_to_output[hidden_node] -= eta * gradient_hidden_to_output
validation_data = []
for i in range(training_count):
data, expected = training_data[i]
for node in range(hidden_layer):
preActivation_H[node] = np.dot(data, weights_input_to_hidden[:, node])
postActivation_H[node] = sigmoid(preActivation_H[node])
preActivation_O = np.dot(postActivation_H, weights_hidden_to_output)
postActivation_O = sigmoid(preActivation_O)
#print(postActivation_O)
if postActivation_O > 0.5:
output = 1
else:
output = 0
validation_data.append((postActivation_O, expected, output))
```
Get help from others!
Recent Questions
Recent Answers
© 2024 TransWikia.com. All rights reserved. Sites we Love: PCI Database, UKBizDB, Menu Kuliner, Sharing RPP