notebook.community

Edit and run



In [1]:

    
import pandas as pd
import numpy as np
import math
import random
%matplotlib inline



In [2]:

    
# 乱数シードを固定する
random.seed(0)



In [3]:

    
# 乱数の係数
random_factor = 0.05
# サイクルあたりのステップ数
steps_per_cycle = 50
# 生成するサイクル数
number_of_cycles = 100



In [4]:

    
df = pd.DataFrame(np.arange(steps_per_cycle * number_of_cycles + 1), columns=["t"])
df.head()



In [5]:

    
random.uniform(-1.0, +1.0) * random_factor









    Out[5]:





0.034442185152504814



In [6]:

    
df["sin_t"] = df.t.apply(lambda x: math.sin(x * (2 * math.pi / steps_per_cycle) + random.uniform(-1.0, +1.0) * random_factor))



In [7]:

    
df[["sin_t"]].plot()









    Out[7]:





<matplotlib.axes._subplots.AxesSubplot at 0x10cd2f320>



In [8]:

    
df[["sin_t"]].head(steps_per_cycle * 2).plot()









    Out[8]:





<matplotlib.axes._subplots.AxesSubplot at 0x10cd58278>



In [9]:

    
df["sin_t+1"] = df["sin_t"].shift(-1)



In [10]:

    
df.tail()



In [11]:

    
df.dropna(inplace=True)
df.tail()



In [12]:

    
df[["sin_t", "sin_t+1"]].head(steps_per_cycle).plot()









    Out[12]:





<matplotlib.axes._subplots.AxesSubplot at 0x10cd5d160>



In [13]:

    
matrix = df[["sin_t", "sin_t+1"]].as_matrix()
matrix









    Out[13]:





array([[ 0.02579258,  0.11744915],
       [ 0.11744915,  0.22526896],
       [ 0.22526896,  0.36917262],
       ..., 
       [-0.32294043, -0.20061926],
       [-0.20061926, -0.14804184],
       [-0.14804184, -0.01545028]])



In [14]:

    
np.save("noised.npy", matrix)



In [ ]:

	t	sin_t	sin_t+1
4996	4996	-0.460640	-0.322940
4997	4997	-0.322940	-0.200619
4998	4998	-0.200619	-0.148042
4999	4999	-0.148042	-0.015450
5000	5000	-0.015450	NaN

	t	sin_t	sin_t+1
4995	4995	-0.551502	-0.460640
4996	4996	-0.460640	-0.322940
4997	4997	-0.322940	-0.200619
4998	4998	-0.200619	-0.148042
4999	4999	-0.148042	-0.015450