Width-basedLookaheadswithLearntBasePolicies and Heuristics OvertheAtari-2600Benchmark