DLFast: Simplifying Deep Learning with Python

dlfast is a Python deep learning library that's revolutionizing the way we build neural networks. Designed for both beginners and experienced data scientists, dlfast empowers users to create intricate deep learning models effortlessly. With its intuitive API and high-level abstractions, you can construct complex networks with just a few lines of code. The library offers modularity, efficiency, and extensive documentation, making it a powerful tool for anyone looking to harness the potential of deep learning.

ANN

The ANN function is a versatile tool for building and training neural networks with ease. It provides options for various tasks, data preprocessing, model architecture, optimization, and more. Whether you're working on binary classification, multiclass classification, or regression, ANN simplifies the process and offers flexibility in customization.

Parameters

• X (numpy.ndarray): The feature data.
• y (numpy.ndarray): The target data.
• task (str, optional): Specifies the task type. Options are 'binary' (default), 'multiclass', or 'regression'.
• scaler (str, optional): Specifies the data scaling method. Options are 'standard' (default), 'robust', or 'minmax'.
• epochs (int, optional): The number of training epochs (default: 10).
• optimizer (str, optional): Specifies the optimizer. Options are 'Adam' (default) or 'SGD'.
• activation (str, optional): Specifies the activation functions for input and hidden layers. Format: 'input_activation/hidden_activation' (default: 'relu').
• layers (tuple, optional): Defines the neural network architecture as a tuple of layer sizes (default: (100, 50, 3)).
• early_stop (bool, optional): Enables or disables early stopping (default: True).
• dropout (float, optional): Dropout rate for regularization (default: None).
• save (bool, optional): Specifies whether to save the model as an h5 file (default: False).

Returns

• model (tensorflow.python.keras.engine.sequential.Sequential): The trained neural network model.
• evaluation_results (float): The evaluation results (e.g., accuracy for classification, mean squared error for regression).

Example Usage

from dlfast import ANN

# Assuming you have your X and y data loaded from your dataset
X, y = load_data()

Binary Classification Example:

model, results = ANN(X, y, task='binary', scaler='standard', epochs=20, optimizer='Adam', activation='relu/sigmoid', early_stop=True, dropout=0.2, save=True)

Multiclass Classification Example:

model, results = ANN(X, y, task='multiclass', scaler='standard', epochs=20, optimizer='Adam', activation='relu/softmax', early_stop=True, dropout=0.2, save=True)

Regression Example:

model, results = ANN(X, y, task='regression', scaler='standard', epochs=20, optimizer='Adam', activation='relu/linear', early_stop=True, dropout=0.2, save=True)