context_{new} = compact(context_full) + β(learning_trajectory)