-
Notifications
You must be signed in to change notification settings - Fork 1
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
neural net doesn't correctly learn Xor #65
Comments
This python code also does something similar therefore bug is in the logic itself. import numpy as np
def randf():
return np.random.rand()
def pow(a, b):
return a ** b
def exit(code):
import sys
sys.exit(code)
def exp(a):
return np.exp(a)
def randmat(r, c):
return np.random.rand(r, c)
def submat(m1, m2):
return np.subtract(m1, m2)
def powmat(m, p):
return np.power(m, p)
def max(a, b):
return np.maximum(a, b)
def relu(m):
return np.maximum(0.0, m)
def sigmoid(m):
return 1.0 / (1.0 + np.exp(-m))
def summat(m):
return np.sum(m)
def matmul(m1, m2):
return np.dot(m1, m2)
def matmulc(mat, v):
return np.multiply(mat, v)
def transpose(m):
return np.transpose(m)
def print_size(m):
print("size: {} x {}".format(len(m), len(m[0])))
def main():
input_size = 2
hidden_size = 36
output_size = 1
w1 = randmat(hidden_size, input_size)
w2 = randmat(output_size, hidden_size)
print("w1:\n", w1)
print("w2:\n", w2)
lr = 0.01
# training xor
_x = np.array([[0.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 1.0]])
x = np.transpose(_x)
y = np.array([[0.0, 1.0, 1.0, 0.0]])
print("x:\n", x)
for i in range(10000):
# forward
h = relu(matmul(w1, x))
h2 = sigmoid(matmul(w2, h))
# loss
loss = summat(powmat(submat(h2, y), 2.0))
if i % 1000 == 0:
print("i: {}".format(i))
print("loss: {}".format(loss))
print("preds:\n", h2)
# backward
grad_h2 = submat(h2, y)
grad_w2 = matmul(grad_h2, transpose(h))
grad_h = matmul(transpose(w2), grad_h2)
grad_w1 = matmul(grad_h, transpose(x))
# update
w1 = submat(w1, matmulc(grad_w1, lr))
w2 = submat(w2, matmulc(grad_w2, lr))
if __name__ == "__main__":
main() |
ammarbinfaisal
changed the title
neuralnet doesn't correctly learns xor
neural net doesn't correctly learn Xor
Nov 24, 2023
Also this python code takes 0.25 seconds and sahl code takes 3.3s :( |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
figure out whats wrong in the code https://github.com/ammarbinfaisal/sahl/blob/master/samples/neuralnet.sahl
The text was updated successfully, but these errors were encountered: