MaxDropout removes most active neurons to prevent overfitting in deep neural networks.
problem Preventing overfitting in deep neural networks.
method MaxDropout removes the most active neurons in each hidden layer to enforce sparsity and prevent overfitting.
result MaxDropout improves neural network performance in image classification and achieves comparable results to existing regularizers.