It has been shown recently that width-based search algorithms can be employed to search over the regression space (backward search). While many benchmarks are challenging for the width-based backward search, it performs significantly better than the forward counterparts in certain domains. This orthogonal behavior of forward and backward width-based search is quite suitable for an integrated approach. Indeed, it has been shown that a simple forward-backward integration that runs forward best-first width search (BFWS) with novelty pruning followed by the backward counterpart results in better coverage than both. Similarly, pairing forward-backward pruned BFWS algorithm with the state-of-the-art Dual-BFWS improves the overall coverage over the IPC satisficing benchmark. In this paper, we present an integration of approximate novelty search with the forward-backward BFWS.