Deep learning also called hierarchical learning is part of a broader family of machine learning method based on the learning data representation .learning can be supervised, unsupervised or reinforcement there are many deep learning libraries nowadays. Libraries contain direct function by which we can import the library and we can directly perform an algorithm on the data .now days number of such type libraries are available with their available feature and benefits .we have to select an appropriate tool for performing a particular task is difficult to decide. This research represents the comparative analysis of deep learning libraries .the main libraries which I will compare is sensor flow, pytorch, theano, and caffe the parameters for comparing the libraries are the adoption, dynamic and static graph definition, debugging, visualization and data parallelism.