In deep learning models, which regularization technique penalizes the squared magnitude of the coefficients?

L2 Regularization, also known as weight decay, penalizes the squared magnitude of the coefficients in deep learning models. It adds a term to the loss function that discourages large weight values, helping to prevent overfitting. By penalizing the magnitude of weights, L2 regularization encourages the model to distribute its learning across many features, resulting in smoother weight values and reducing the risk of overfitting.