nn.py

gistfile1.txt

"""
Test your network here!

No need to change this code, but feel free to tweak it
to test your network!

Make your changes to backward method of the Sigmoid class in miniflow.py
"""

import numpy as np
from miniflow import *

X, W, b = Input(), Input(), Input()
y = Input()
f = Linear(X, W, b)
a = Sigmoid(f)
cost = MSE(y, a)

X_ = np.array([[1., 2.], [3., 4.]])
n_rows = X_.shape[0]
n_features = X_.shape[1]
n_hidden = 1
# W_ = np.array([[2.], [3.]])
W_ = np.random.randn(n_features, n_hidden) 
# b_ = np.array([5.])
b_ = np.zeros(1)

y_ = np.array([12, 20])

'''
X_ = np.array([[-1., -2.], [-1, -2]])
W_ = np.array([[2.], [3.]])
b_ = np.array([-3.])
y_ = np.array([1, 2])
'''

feed_dict = {
    X: X_,
    y: y_,
    W: W_,
    b: b_,
}
trainables = [W, b]

graph = topological_sort(feed_dict)

# forward_and_backward(graph)
# return the gradients for each Input
# gradients = [t.gradients[t] for t in [X, y, W, b]]

epochs = 10

for i in range(epochs):
    loss = 0
    for j in X.value:
        # Step 2
        forward_and_backward(graph)
        
        # Step 3
        sgd_update(trainables, .1)
        
        loss += graph[-1].value
        
    print("Epoch: {}, Loss: {:.3f}".format(i+1, loss/n_rows))