apollo-ad 0.0.3

Auto Differentiation Tools

cs107-FinalProject, Group 16

Group members	Email
Connor Capitolo	[e-mail]
Kexin Huang	[e-mail]
Haoxin Li	[e-mail]
Chen Lucy Zhang	[e-mail]

Installation (we assume you are familiar with virtual environments and Git)

# install virtualenv
pip install virtualenv

# create virtual environment
virtualenv apollo_ad_env

# activate virtual environment
source apollo_ad_env/bin/activate

# clone from GitHub
git clone https://github.com/West-Coast-Quaranteam/cs107-FinalProject.git

# get into the folder
cd cs107-FinalProject

# install requirements
pip install -r requirements.txt

# Test the package
# From directory apollo_ad/tests/ run the module test.py
pytest test_scalar.py

Examples

We show two examples here to use apollo_ad.

First, to calculate the derivative of y = cos(x) + x^2 at x = 2:

from apollo_ad.apollo_ad import *
x = Variable(2)
y = Variable.cos(x) + x ** 2

print(y) # Value: 3.5838531634528574 , Der: [3.09070257]
print(y.var) # 3.583853163452857
print(y.der) # [3.09070257]

assert y.var == np.cos(2) + 4
assert y.der == -np.sin(2) + 2 * 2

Second, to calculate the derivative of y = 2 * log(x) - sqrt(x) / 3 at x = 2:

from apollo_ad.apollo_ad import *
x = Variable(2) 
y = 2 * Variable.log(x) - Variable.sqrt(x)/3

print(y) # Value: 0.9148898403288588 , Der: [0.88214887]
print(y.var) # 0.9148898403288588
print(y.der) # [0.88214887]

assert np.around(y.var, 4) == np.around(2 * np.log(2) - np.sqrt(2)/3, 4)
assert np.around(y.der, 4) == np.around(2 * 1/2 - 1/3 * 1/2 * 2**(-1/2), 4) 

You can also run the above examples by:

python demo.py