The well-known East-West Challenge is originally an ILP problem. The problem consist of predicting whether a train is eastbound or westbound. A train (bag) contains a variable number of cars (instances) that have different shapes and carry different loads (instance-level attributes). As the standard MI assumption is asymmetric and it is not clear whether an eastbound train or a westbound train can be regarded as a positive example in the MI setting, we consider two MI versions of the data for our experiments. This dataset contains eastbound trains as positive examples.