Which of the following are regularization techniques?

Regularization techniques are methods used to prevent overfitting in a machine learning model by adding a penalty term to the loss function. Here are some commonly used regularization techniques:

1. L1 Regularization (Lasso Regression): This technique adds an absolute value of magnitude of coefficient as penalty term to the loss function.

2. L2 Regularization (Ridge Regression): This technique adds the squared magnitude of coefficient as penalty term to the loss function.

3. Elastic Net: This is a middle ground between Lasso Regression and Ridge Regression. It adds both absolute and squared magnitude of coefficient as penalty term to the loss function.

4. Dropout: This is a regularization technique used in neural networks. It randomly drops out (by setting the activation to zero) each neuron in a layer with a certain probability during training.

5. Early Stopping: In this method, we stop the training before the model starts to overfit.

6. Data Augmentation: This technique increases the amount of training data using information only in the training data, by creating transformed versions of the input data.

7. Batch Normalization: This technique allows each layer of a network to learn by itself a little bit more independently of other layers.

8. Noise Injection: This technique involves adding a small amount of noise to the input data or to the output layer during training.

9. Weight Decay: This is a very common form of regularization, where a penalty is added to the loss function that is proportional to the size of the weights.

10. Max-Norm Constraints: This technique involves constraining the weights of the incoming connections to each neuron in the hidden layer to have a maximum norm.

Remember, the choice of regularization technique depends on the specific problem and the type of model being used.