Implementing Recurrent Neural Network using Numpy

Recurrent neural network (RNN) is one of the earliest neural networks that was able to provide a break through in the field of NLP. The beauty of this network is its capacity to store memory of previous sequences due to which they are widely used for time series tasks as well. High level frameworks like Tensorflow and PyTorch abstract the mathematics behind these neural networks making it difficult for any AI enthusiast to code a deep learning architecture with right knowledge of parameters and layers. In order to resolve these type of inefficiencies the mathematical knowledge behind these networks is necessary. Coding these algorithms from scratch gives an extra edge by helping AI enthusiast understand the different notations in research papers and implement them in practicality. If you are new to the concept of RNN please refer to MIT 6.S191 course, which is one of the best lectures giving a good intuitive understanding on how RNN work.