Which method applies a transformation to keep output close to a mean of 0 and standard deviation of 1?

Batch Normalization is a technique used during training of deep learning models to improve convergence speed and overall performance by addressing the internal covariate shift. It applies a transformation to the input of each layer by standardizing the activations to have a mean of 0 and a standard deviation of 1. This is accomplished by normalizing the input batch within the network as it's being processed.

The process involves calculating the mean and standard deviation for each feature in the batch, then using these statistics to adjust the activations. This helps stabilize the learning process and allows for higher learning rates, which can lead to faster convergence and potentially better overall model performance.

In contrast, other options like Dropout, Normalization, and L2 Regularization serve different purposes. Dropout is used as a regularization technique to prevent overfitting by randomly setting a fraction of the neurons to zero during training. Normalization, while it can refer to various processes to adjust feature scales, does not specifically encompass the batch-based transformation focused on maintaining zero mean and unit variance in the context of neural network layers. L2 Regularization, meanwhile, discourages large weights by adding a penalty based on the norm of the weights, but it does not apply a transformation to keep the output