A Survey on Entropy Optimized Feature-based Bag-of-Words Representation for Information Retrieval

In this paper, we present a supervised dictionary learning method for improving the component based Bag-of-Words (BoW) representation towards Information Retrieval. Taking after the bunch theory, which expresses that focuses in a similar group are probably going to satisfy a similar data require, we propose the utilization of an entropy-based enhancement basis that is more qualified for recovery of order. We show the capacity of the proposed strategy, curtailed as EO-BoW, to enhance the recovery execution by giving broad analyses on two multi-class picture datasets. The BoW model can be connected to different spaces too, so we additionally assess our approach utilizing a gathering of 45 time-arrangement datasets, a content dataset and a video dataset. The increases are three-crease since the EO-BoW can enhance the mean Average Precision, while decreasing the encoding time and the database stockpiling necessities. At long last, we give prove that the EO-BoW keeps up its representation capacity notwithstanding when used to recover objects from classes that were not seen amid the preparation.