my-nn-framework

My implementation of Neural Network for better understanding

• almost same usage with keras

• sadly, sigmoid only as activation function

• sadly, sgd only as opt

• sadly, pay no attention to parameter init

Setup

1. create & activate virtualenv

2. install lib by

   pip3 install -r requirements.txt
   

Usage

Example

from my_nn_framework import Model, Dense


model = Model(
    Dense("sigmoid", in_features=64, out_features=30),
    Dense("sigmoid", out_features=10),
)

model.compile(optimizer="sgd", loss="mse")

model.fit(
    x_train,
    y_train,
    epochs=300,
    batch_size=32,
    learning_rate=0.01
)

predicted = model.predict(x_test)

accuracy = Model.evaluate(predicted, y_test)

>>> model.summary()
<bound method Model.summary of Model with 2 layers
    ([Dense: input_features=64, out_features=30],
    [Dense: input_features=30, out_features=10])
>

After Model.fit(), loss graph like

would be saved as out.png

Acknowledgements

• Neural Networks and Deep Learning