def test_forward1(self):
x = Variable(np.array([[1, 2, 3], [4, 5, 6]]))
w = Variable(x.data.T)
y = F.matmul(x, w)
res = y.data
expected = np.array([[14, 32], [32, 77]])
self.assertTrue(array_allclose(res, expected))
def test_matmul(self):
x = Variable(np.random.randn(2, 3))
W = Variable(np.random.randn(3, 4))
y = F.matmul(x, W)
y.backward()
self.assertEqual(x.grad.shape, (2, 3))
self.assertEqual(W.grad.shape, (3, 4))
def test_matmul(self):
x = Variable(np.array([[1, 1], [1, 1]]))
W = Variable(np.array([[2], [2]]))
y = matmul(x, W)
y.backward()
assert_array_equal(y.data, np.array([[4], [4]]))
assert_array_equal(x.grad.data, np.array([[2, 2], [2, 2]]))
assert_array_equal(W.grad.data, np.array([[2], [2]]))
if '__file__' in globals():
import os, sys
sys.path.append(os.path.join(os.path.dirname(__file__), '..'))
import numpy as np
from dezero import Variable
import dezero.functions as F
from dezero.utils import sum_to
# ベクトルの内積
a = np.array([1, 2, 3])
b = np.array([4, 5, 6])
c = np.dot(a, b)
print(c)
# 行列の積
a = np.array([[1, 2], [3, 4]])
b = np.array([[5, 6], [7, 8]])
c = np.dot(a, b)
print(c)
x = Variable(np.random.randn(2, 3))
W = Variable(np.random.randn(3, 4))
y = F.matmul(x, W)
y.backward()
print(x.grad.shape)
print(W.grad.shape)
def predict(x):
y = F.matmul(x, W) + b
return y
import os, sys sys.path.append(os.path.join(os.path.dirname(__file__), '..')) import numpy as np from dezero import Variable import dezero.functions as F x = Variable(np.random.randn(2, 3)) w = Variable(np.random.randn(3, 4)) t = F.matmul(x, w) y = F.sum(t) y.backward() print(x.grad.shape) print(w.grad.shape)
def test_backward(self):
x = Variable(np.arange(4).reshape(2, 2))
W = Variable(np.arange(6).reshape(2, 3))
y = matmul(x, W)
y.backward()
def test_forward(self):
x = Variable(np.arange(4).reshape(2, 2))
W = Variable(np.arange(6).reshape(2, 3))
y = matmul(x, W)
assert_equal(y.data, np.array([[3, 4, 5], [9, 14, 19]]))
def predict(x):
y = F.matmul(x, w1) + b1
y = F.sigmoid(y)
y = F.matmul(y, w2) + b2
return y
def test_backward2(self):
x_data = np.random.randn(10, 1)
w_data = np.random.randn(1, 5)
f = lambda w: F.matmul(Variable(x_data), w)
self.assertTrue(gradient_check(f, w_data))
def test_backward1(self):
x = np.random.randn(3, 2)
w = np.random.randn(2, 3)
f = lambda x: F.matmul(x, Variable(w))
self.assertTrue(gradient_check(f, x))
def gram_mat(x):
N, C, H, W = x.shape
features = x.reshape(C, -1)
gram = F.matmul(features, features.T)
return gram.reshape(1, C, C)
def predict(x):
return F.matmul(x, W) + b
import numpy as np from dezero import Variable import dezero.functions as F # Inner products a = np.array([1, 2, 3]) b = np.array([4, 5, 6]) a, b = Variable(a), Variable(b) # Optional c = F.matmul(a, b) print(c) # Matrix product a = np.array([[1, 2], [3, 4]]) b = np.array([[5, 6], [7, 8]]) c = F.matmul(a, b) print(c)
y = F.matmul(x, W) + b
return y
def mean_squared_error(x0, x1):
diff = x0 - x1
return F.sum(diff**2) / len(diff)
lr = 0.1
iters = 100
for i in range(iters):
y_pred = predict(x)
loss = mean_squared_error(y, y_pred)
W.clear_grad()
b.clear_grad()
loss.backward()
W.data -= lr * W.grad.data
b.data -= lr * b.grad.data
print(W, b, loss)
from IPython import embed
embed()
sorted_x = sorted(x.data.squeeze())
predictions = sorted((F.matmul(x, W) + b).data.squeeze())
plt.scatter(x.data.squeeze(), y.data.squeeze())
plt.plot(sorted_x, predictions, color='r')
plt.show()
