CNN-python

The implementation of a Convolutional neural network based on Python-NumPy.

Layers

• Fully connected layer (forward & backward)
• Convolutional layer (forward & backward)
• Pooling layer (forward & backward)
• Zero Padding layer (forward & backward)
• Dropout (forward & backward)
• [] BatchNorm (forward & backward)

Activate functions

• ReLU (forward & backward)
• Softmax (forward & backward)
• Sigmoid (forward & backward)

Loss functions

• Cross Entropy Loss

Data set

THE MNIST DATABASE of handwritten digits.

Training

• training the CNN.

python main.py