Hands-On Machine Learning on Google Cloud Platform

Implementing L1 regularization

L1 regularization works in a similar way as that of L2; however, the cost function for L1 regularization is different than L2 regularization, as follows:

Note that, in the preceding equation, all the terms remain the same; just the regularization term is the summation of absolute values of weights than squared values of weights.

Let's implement the L1 regularization in code; now we see the corresponding outputs as follows:

Note that, as the regularization term does not involve squaring in L1 regularization, we ...