Created
July 11, 2019 16:43
-
-
Save emuccino/8c3c4b8decbf2721e258262973703358 to your computer and use it in GitHub Desktop.
Training Recurrent Neural Networks on Long Sequences
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import numpy as np | |
#function for getting coordinates given an angle | |
def get_cartesian_coords(nums): | |
theta = nums | |
x = np.cos(theta) | |
y = np.sin(theta) | |
return [x,y] | |
#generate random starting coordinate | |
angle = np.random.uniform(-np.pi, np.pi) | |
coordinates = get_cartesian_coords(random_angle) | |
amplitudes = [abs(coord) for coord in coordinates] | |
random_walk = [amplitudes] | |
for _ in range(9999): | |
#generate random small change in angle | |
direction = 0.01 * np.random.uniform(-1, 1) | |
angle += direction | |
coordinates = get_cartesian_coords(angle) | |
amplitudes = [abs(coord) for coord in coordinates] | |
random_walk.append(amplitudes) | |
timestep = list(range(20000)) | |
#create sequence using generated amplitudes | |
sequence = list(map(lambda i: np.round((((random_walk[i][0])*-1)*np.sin((0.0002)*i)) + \ | |
((random_walk[i][1])*np.sin(0.0002*i)))+ 100 + np.random.normal(0, 0.01),decimals=2),timestep) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment