36 lines
745 B
Python
36 lines
745 B
Python
|
import numpy as np
|
||
|
|
import theano
|
||
|
|
from theano import tensor as T
|
||
|
|
|
||
|
|
theano.config.profile_memory = True
|
||
|
|
theano.config.profile = True
|
||
|
|
|
||
|
|
|
||
|
|
SLOPE = 3.1
|
||
|
|
INTERCEPT = -1.2
|
||
|
|
|
||
|
|
|
||
|
|
def residual_stats_theano(x, y):
|
||
|
|
expected = SLOPE * x + INTERCEPT
|
||
|
|
residuals = y - expected
|
||
|
|
return residuals.mean(), residuals.std()
|
||
|
|
|
||
|
|
|
||
|
|
x_var = T.vector()
|
||
|
|
y_var = T.vector()
|
||
|
|
|
||
|
|
residual_stats = theano.function(
|
||
|
|
inputs=[x_var, y_var],
|
||
|
|
outputs=residual_stats_theano(x_var, y_var),
|
||
|
|
allow_input_downcast=True,
|
||
|
|
profile=True,
|
||
|
|
)
|
||
|
|
|
||
|
|
|
||
|
|
if __name__ == '__main__':
|
||
|
|
x = np.linspace(-10, 10, 1000)
|
||
|
|
y = SLOPE * x + INTERCEPT
|
||
|
|
y += np.random.normal(loc=0.1, scale=0.5, size=x.shape)
|
||
|
|
mn, std = residual_stats(x, y)
|
||
|
|
print('Residual mean=', mn, ', std=', std)
|