Large-Scale Correlation-Based Semantic Classification Using Mapreduce

Nevertheless, even though MCA has shown its eectiveness in CBMIR, its direct application to big data is not scalable. To train an MCA classi cation model, it is required to manipulate large matrices extractedfrom the training data, which impedes the useful utilization of MCA in today's pervasive big data environments. Existing works that utilize MCA for CBMIR tasks do not take into account scalability problems that arise when processing large amounts of data nor provide a framework that eectively utilizes multiple computers to speed up processing. e pertinent question is then how to improve the scalability of MCA and bring it onto the big data scale.