Skip to content

Instantly share code, notes, and snippets.

@alopatindev

alopatindev/gradient_descent.py

Created May 22, 2017 10:42
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save alopatindev/fb59b6216b363a70a1611bc46700d686 to your computer and use it in GitHub Desktop.
Save alopatindev/fb59b6216b363a70a1611bc46700d686 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
gradient_descent.py
#!/usr/bin/env python2
# https://www.youtube.com/watch?v=tIpKfDc295M
# https://www.youtube.com/watch?v=umAeJ7LMCfU
import numpy as np
f = lambda x, y: x * x + 2.0 * y
d_dx = lambda x, y: 2.0 * x
d_dy = lambda x, y: 2.0
gradient = lambda x, y: np.array([d_dx(x, y), d_dy(x, y)])
mu = 0.00005
num_iterations = 100000
xy = np.array([5.0, 10.0])
for _ in xrange(num_iterations):
xy -= mu * gradient(*xy)
print xy, f(*xy)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment