[텐서플로우 정리] 08. 초기값과 난수

이번 글에서는 변수의 초기값을 주는 함수들에 대해 살펴 본다.
이번 코드 또한 functions.py 파일에 있는 함수를 이용해서 결과를 출력하고 있다. (functions.py)


코딩할 때마다 나오는 개념인데.. 매번 헷갈려서 정리한다. uniform이란 단어의 뜻을 이제서야 알았다.
정규분포(normal distribution) - 통계 확률분포에서 가장 중요한 분포로 종 모양(bell shpae)으로 표현되는 확률
균등(균일)분포(uniform distribution) - 각각의 구간에서 동일한 확률로 표현되는 분포로 사각형 모양
잘린(절단)정규분포(truncated normal distribution) - 정규분포에서 일부 구간을 잘라낸 분포
나무위키(정규분포)
나부랭이의 수학블로그(균등분포 개념정리)
잘린정규분포   truncated normal distribution(위키피디아, 영어)


tf.random_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)
Outputs random values from a normal distribution.

Args:
shape: A 1-D integer Tensor or Python array. The shape of the output tensor.
mean: A 0-D Tensor or Python value of type dtype. The mean of the normal distribution.
stddev: A 0-D Tensor or Python value of type dtype. The standard deviation of the normal distribution.
dtype: The type of the output.
seed: A Python integer. Used to create a random seed for the distribution. See set_random_seed for behavior.
name: A name for the operation (optional).

Returns:
A tensor of the specified shape filled with random normal values.


tf.truncated_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)
Outputs random values from a truncated normal distribution.

The generated values follow a normal distribution with specified mean and standard deviation, except that values whose magnitude is more than 2 standard deviations from the mean are dropped and re-picked.

Args:
shape: A 1-D integer Tensor or Python array. The shape of the output tensor.
mean: A 0-D Tensor or Python value of type dtype. The mean of the truncated normal distribution.
stddev: A 0-D Tensor or Python value of type dtype. The standard deviation of the truncated normal distribution.
dtype: The type of the output.
seed: A Python integer. Used to create a random seed for the distribution. See set_random_seed for behavior.
name: A name for the operation (optional).

Returns:
A tensor of the specified shape filled with random truncated normal values.


tf.random_uniform(shape, minval=0, maxval=None, dtype=tf.float32, seed=None, name=None)
Outputs random values from a uniform distribution.

The generated values follow a uniform distribution in the range [minval, maxval). The lower bound minval is included in the range, while the upper bound maxval is excluded.

For floats, the default range is [0, 1). For ints, at least maxval must be specified explicitly.

In the integer case, the random integers are slightly biased unless maxval - minval is an exact power of two. The bias is small for values of maxval - minval significantly smaller than the range of the output (either 2**32 or 2**64).

Args:
shape: A 1-D integer Tensor or Python array. The shape of the output tensor.
minval: A 0-D Tensor or Python value of type dtype. The lower bound on the range of random values to generate. Defaults to 0.
maxval: A 0-D Tensor or Python value of type dtype. The upper bound on the range of random values to generate. Defaults to 1 if dtype is floating point.
dtype: The type of the output: float32, float64, int32, or int64.
seed: A Python integer. Used to create a random seed for the distribution. See set_random_seed for behavior.
name: A name for the operation (optional).

Returns:
A tensor of the specified shape filled with random uniform values.

Raises:
ValueError: If dtype is integral and maxval is not specified.


import tensorflow as tf
from functions import showOperation as showOp

showOp(tf.zeros([2,3])) # [[ 0. 0. 0.] [ 0. 0. 0.]]
showOp(tf.ones([2,3], tf.int32)) # [[1 1 1] [1 1 1]]
showOp(tf.zeros_like(tf.ones([2,3]))) # [[ 0. 0. 0.] [ 0. 0. 0.]]
showOp(tf.fill([2,3], 2)) # [[2 2 2] [2 2 2]]
showOp(tf.fill([2,3], 2.0)) # [[ 2. 2. 2.] [ 2. 2. 2.]]
print('# ------------------------------------------------ #')
showOp(tf.linspace(1.0, 10.0, 4)) # [ 1. 4. 7. 10.]
showOp(tf.range(5)) # [0 1 2 3 4]
showOp(tf.range(0, 5)) # [0 1 2 3 4]
showOp(tf.range(0, 10, 2)) # [0 2 4 6 8]
print('# ------------------------------------------------ #')
# tf.random_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)
# [[-0.14479451 -0.50265551 1.38471031] [-2.46794224 0.56639165 -0.59352636]]
showOp(tf.random_normal([2, 3]))
# [[ 5.0084033 6.28203726 5.49111032] [ 4.81292725 4.8216362 4.82326126]]
showOp(tf.random_normal([2, 3], mean=5.0))
# [[ 0.02739749 0.15692481 -0.18835409] [-0.20757729 0.76803416 -0.07633832]]
showOp(tf.random_normal([2, 3], stddev=0.35))
# [[ 4.7160387 5.51960945 5.0228653 ] [ 4.14505386 5.03473711 5.20692873]]
showOp(tf.random_normal([2, 3], mean=5.0, stddev=0.35, seed=1))
print('# ------------------------------------------------ #')
# -0.1293 -0.0633 0.0984 0.0508 0.1916 0.1197 -0.3135 -0.0823 -0.0300 0.0430
for _ in range(10):
showOp(tf.reduce_sum(tf.random_normal([2, 3], stddev=0.35) / 6))
# 5.03024 4.9332 4.97116 4.96042 4.99974 5.15176 4.961 4.64518 5.26943 5.18463
for _ in range(10):
showOp(tf.reduce_sum(tf.random_normal([2, 3], mean=5.0, stddev=0.35) / 6))
print('# ------------------------------------------------ #')
# tf.random_uniform(shape, minval=0, maxval=None, dtype=tf.float32, seed=None, name=None)
# [[ 0.67391849 0.66735017 0.0794853 ] [ 0.64219582 0.47089899 0.68388402]]
showOp(tf.random_uniform([2, 3]))
# [[ 1.28794432 2.02983379 2.73823977] [ 2.97947931 4.48065186 4.69495058]]
showOp(tf.random_uniform([2, 3], minval=5)) # no error. do not work.
# [[ 13.32174492 3.7804091 10.80069256] [ 8.99943733 0.50998271 9.10907364]]
showOp(tf.random_uniform([2, 3], maxval=15))
# [[ 8.30814171 8.18294716 6.43582296] [ 5.31517982 9.81415558 8.0894165 ]]
showOp(tf.random_uniform([2, 3], 5, 10))
# [[ 7.41970491 7.5310154 6.27571869] [ 5.62371159 9.66505051 8.32304668]]
showOp(tf.random_uniform([2, 3], 5, 10, seed=7))
print('# ------------------------------------------------ #')
# tf.random_shuffle(value, seed=None, name=None)
showOp(tf.random_shuffle(tf.Variable([1,2,3,4,5,6]))) # [3 4 6 5 2 1]
showOp(tf.random_shuffle(tf.Variable([[1,2], [3,4], [5,6]]))) # [[5 6] [3 4] [1 2]]
print('# ------------------------------------------------ #')
# [[ 0.23903739 0.92039955 0.05051243] [ 0.49574447 0.83552229 0.02647042]]
showOp(tf.random_uniform([2, 3], seed=1))
# [[ 0.23903739 0.92039955 0.05051243] [ 0.49574447 0.83552229 0.02647042]]
showOp(tf.random_uniform([2, 3], seed=1))
print('# ------------------------------------------------ #')
# tf.truncated_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)
# The generated values follow a normal distribution with specified mean and standard deviation,
# except that values whose magnitude is more than 2 standard deviations
# from the mean are dropped and re-picked.
# [[-0.50039721 -1.03818107 -0.1909811 ] [-0.41344216 -0.91877717 0.24455361]] showOp(tf.truncated_normal([2,3])) # [[ 3.40512562 -0.11788981 1.92399371] [ 6.6576376 4.21847486 4.41285849]] showOp(tf.truncated_normal([2,3], stddev=3.5))