Continuous learning in single-incremental-task scenarios