NNLib

NNLib can be used to create neural network models. This module does not use Tensorflow or other ai libraries because this module aims to understand background mathematics of a neural network model and easily make hyperparameter tuning.

Requirements

• Python 3.x
• Numpy
• Pandas
• Matplotlib
• Sklearn.metrics

Example

model = NNLib()

x_train, y_train, x_test, y_test = model.get_train_test_data(train_ratio= 70, data = df, label = "label", return_np_array= 1)

x_train_norm = model.apply_normalization(data= x_train,normalization="gaussian")

x_test_norm = model.apply_normalization(data= x_test,normalization="gaussian")

y_train_one_hot_encoding = model.apply_one_hot_encoding(y_train)

model.create_network(x_train_norm, parameters_initilization_method="He",learning_rate = 0.01,loss_function="cross_entropy")

model.add_layer(64, activation_function = "Relu")

model.add_layer(32, activation_function = "Relu")

model.add_layer(16, activation_function = "Relu")

model.add_layer(2, activation_function = "softmax")

model.train(x_train_norm, y_train_one_hot_encoding, mini_batch_size=10)

predictions = model.predict(x_test_norm)

model.evaluate(y_pred=predictions,y_true=y_test)