Model Fusion-Based Batch Learning with Application to Oil Spills Detection

Data split into batches is very common in real-world applications. In speech recognition and handwriting identification, the batches are different people. In areas like oil spill detection and train wheel failure prediction, the batches are the particular circumstances when the readings were recorded. The recent research has proved that it is important to respect the batch structure when learning models for batched data. We believe that such a batch structure is also an opportunity that can be exploited in the learning process. In this short paper, we investigated the novel method for dealing with the batched data. We applied the developed batch learning techniques to detect oil spills using radar images collected from satellite stations. This paper reports some progress on the proposed batch learning method and the preliminary results obtained from oil spills detection.