We present a new method which solves a double-barelled LASSO in a convex least squares approach. In the presented method we focus on the scenario where one is interested in (or limited to) a primal (feature) representation for the first view while having a dual (kernel) representation for the second view. DB-LASSO minimises the number of features used in both the primal and dual projections while minimising the error (maximising the correlation) between the two views.