Fully Connected Neural Network

Requirements

Only Numpy !

RUN

git clone https://github.com/MarvinMartin24/Fully-Connected-Neural-Network.git

Go to your file directory, and run this command :

python3 main.py

Usage

net = Network()
net.add(FCLayer(input_size, nb_neurone))
net.add(ActivationLayer(activation, activation_prime))
net.add(FCLayer(prev_nb_neurone, output_size))
net.add(ActivationLayer(activation, activation_prime))
net.use(Loss, Loss_prime)
net.fit(x_train, y_train, Epoch, learning_rate)
out = net.predict(x_train)
print(out)

Implementation : XOR

• Input / Data:

x_train = np.array([[[0,0]], [[0,1]], [[1,0]], [[1,1]]])
y_train = np.array([[[0]], [[1]], [[1]], [[0]]])

• Output / Prediction :

  Input =
  [[[0 0]]
  [[0 1]]
  [[1 0]]
  [[1 1]]]
  Prediction =
  [[[0.]]
  [[1.]]
  [[1.]]
  [[0.]]]

  Understand the Maths

  This code is based on this Medium

Colaborators

Omar Aflak (Author of the Medium Article)
and Marvin Martin