Robust linear regression using Least Absolute Deviations with L1 regularization in Tensorflow. The optimization problem is solved by applying the Majorization-Minimization (MM) framework, which generates a sequence of estimators for the linear weights which converges to the maximum a posteriori estimator.