PureMLP
is a minimal, customizable feed-forward neural network implemented using only NumPy.
It supports configurable layer sizes, activation functions, forward propagation, backpropagation, and mini-batch training with gradient descent.
It also includes a basic testing and visualization function for classification tasks.
NumPy
Pandas
Matplotlib
from PureMLP import MLP
model = MLP(layers=[784, 64, 10], activation_functions=['relu', 'softmax'])
model.train(X_train, Y_train, epochs=10, learning_rate=0.01, batch_size=64)
predictions = model.predict(X_test)
model.test(X_sample, true_label)