| | |
- apply_dropout_jit(X, dropout_rate)
- Apply dropout to activation values.
- calculate_bce_with_logits_loss(logits, targets)
- Calculate binary cross-entropy loss with logits.
- calculate_cross_entropy_loss(logits, targets)
- Calculate cross-entropy loss for multi-class classification.
- calculate_huber_loss(y_pred, y_true, delta=1.0)
- Helper function to calculate the Huber loss. Handles 1D and 2D inputs.
- calculate_loss_from_outputs_binary(outputs, y, weights, reg_lambda)
- Calculate binary classification loss with L2 regularization.
- calculate_loss_from_outputs_multi(outputs, y, weights, reg_lambda)
- Calculate multi-class classification loss with L2 regularization.
- calculate_mae_loss(y_pred, y_true)
- Helper function to calculate the mean absolute error loss. Handles 1D and 2D inputs.
- calculate_mse_loss(y_pred, y_true)
- Helper function to calculate the mean squared error loss. Handles 1D and 2D inputs.
- compute_l2_reg(weights)
- Compute L2 regularization for weights.
- evaluate_batch(y_hat, y_true, is_binary)
- Evaluate accuracy for a batch of predictions.
- evaluate_jit(y_hat, y_true, is_binary)
- Evaluate model performance and return accuracy and predictions.
- evaluate_regression_jit(y_pred, y_true, loss_function)
- Evaluate model performance for regression tasks using Numba.
Args:
y_pred (ndarray): Model predictions.
y_true (ndarray): True target values.
loss_function (object): The JIT loss function instance (e.g., JITMeanSquaredErrorLoss).
Returns:
tuple: Metric value (e.g., MSE) and the predictions.
- leaky_relu(z, alpha=0.01)
- Apply Leaky ReLU activation function.
- leaky_relu_derivative(z, alpha=0.01)
- Compute the derivative of the Leaky ReLU activation function.
- one_hot_encode(y, num_classes)
- One-hot encode a vector of class labels.
- process_batches_binary(X_shuffled, y_shuffled, batch_size, layers, dropout_rate, dropout_layer_indices, reg_lambda, dWs_acc, dbs_acc)
- Process batches for binary classification.
- process_batches_multi(X_shuffled, y_shuffled, batch_size, layers, dropout_rate, dropout_layer_indices, reg_lambda, dWs_acc, dbs_acc)
- Process batches for multi-class classification.
- process_batches_regression_jit(X_shuffled, y_shuffled, batch_size, layers, dropout_rate, dropout_layer_indices, reg_lambda, dWs_acc, dbs_acc, loss_calculator_func)
- Process batches for regression tasks using Numba.
- relu(z)
- Apply ReLU activation function.
- relu_derivative(z)
- Compute the derivative of the ReLU activation function.
- sigmoid(z)
- Apply sigmoid activation function.
- sigmoid_derivative(z)
- Compute the derivative of the sigmoid activation function.
- softmax(z)
- Apply softmax activation function.
- sum_axis0(arr)
- Sum elements along axis 0.
- sum_reduce(arr)
- Sum elements along the last axis and reduce the array.
- tanh(z)
- Apply tanh activation function.
- tanh_derivative(z)
- Compute the derivative of the tanh activation function.
|