An $L\sb 1$-norm procedure for fitting overlapping clustering models to proximity data