Parallel Approach for Document Representation using Dictionary Learning

There has been drastic advancements in data science both in terms of algorithm development and storage is concerned. This work takes into consideration an area which is coming of use in data handling based on sparse representation. The collection of documents which is considered as a dictionary is large with a massive amount of data providing in. So in order to represent an incoming document in sparse mode, we have considered parallelizing the algorithm for dictionary learning and sparse code generation. This work show good improvement in reducing the time for generating sparse representation with help of GPUs. There is a 2x increase in time when compared to CPU usage for generating the same sparse code for an input collection.