原帖是 Tiezhen 在做创新工场 workshop 活动时向参会者提出的问题，很有意义，转发如下：
请依次完成以下任务，并回复本帖。
Test 0
import tensorflow as tf
import sys
print (tf.__version__)
print (sys.version)
# Question: What are the outputs and why?
Test 1.0
import tensorflow as tf
a = tf.constant (1)
b = tf.constant (1)
c = tf.add (a, b)
print (c)
with tf.Session () as sess:
print (sess.run (c))
# Question: What's the output?
Test 1.1
import tensorflow as tf
# 2-D tensor `a`
# [[1, 2, 3],
# [4, 5, 6]]
a = tf.constant ([1, 2, 3, 4, 5, 6], shape=[2, 3])
# 2-D tensor `b`
# [[ 7, 8],
# [ 9, 10],
# [11, 12]]
b = tf.constant ([7, 8, 9, 10, 11, 12], shape=[3, 2])
# Question: write a program to calculate the matrix mautiplication of a and b (i.e. a * b) # and what's the output?
# API doc is available here: https://tensorflow.google.cn/api_docs/python/tf/
# YOUR CODE HERE
Test 1.2
import tensorflow as tf
x = tf.get_variable ('x', shape=[1], initializer=tf.constant_initializer (3))
y = tf.square (x)
y_grad = tf.gradients (y, [x])
init = tf.global_variables_initializer ()
with tf.Session () as sess:
sess.run (init)
print (sess.run (y_grad))
# Question: what's the output and what does the output mean?
Test 1.3
import tensorflow as tf
f = lambda x: 3*x
x_init = 1
y_expected = 5
def find_root (f, x_init, y_expected, learning_rate):
x = tf.get_variable ('x', shape=[1], initializer=tf.constant_initializer (x_init))
init = tf.global_variables_initializer ()
with tf.Session () as sess:
sess.run (init)
for _ in range (EPOCHS):
y_predict = ...
loss = ...
y_grad = tf.gradients (loss, [x])
op = tf.assign_sub (x, ...)
op = tf.Print (op, [...], "updates: ")
sess.run (op)
return sess.run (x)
# Question: write a function to calculate x so that f (x) = y_expected using gradient descent method
# Question: Try to change the learning rate and explain what you found
# Question: How to reduce the number of epoch?
Test 2
# Question: Write a XOR model to prevent the result of a XOR b
# Question: Write your own metrics, loss and optimizer
import tensorflow as tf
import numpy as np
X = np.array ([[0,0],[0,1],[1,0],[1,1]])
y = np.array ([[0],[1],[1],[0]])
class XOR (tf.keras.Model):
def __init__(self):
super (XOR, self).__init__(self)
self.dense1 = ...
self.dense2 = ...
def call (self, inputs):
x = self.dense1 (inputs)
x = self.dense2 (x)
return x
def accuracy (y_true, y_pred):
return UnImplemented
def loss (y_true, y_pred):
return UnImplemented
class Optimizer (tf.keras.optimizers.Optimizer):
def __init__(self, learning_rate):
super (Optimizer, self).__init__()
self._learning_rate = learning_rate
...
def get_updates (self, loss, params):
updates = []
# TODO: Implement here
return updates
def run ():
m = XOR ()
m.compile (loss=loss, optimizer=Optimizer (...), metrics=[accuracy])
m.fit (X, y, epochs=EPOCHS)
print (m.predict (X))
if __name__ == "__main__":
run ()