I have a simple mini-batch fed RNN written in python (numpy) that I can't get to function correctly. The data sets are tiny with only 15-20 examples and some data sets converge while others don't. With how tiny the data sets are; I can tell there is a deeper issue with the code. Anyone with an above average knowledge of recurrent neural networks can probably find the solution fairly quickly. I think it's with how I am handling the accumulating mini-batch deltas. The previous sentence should indicate if your understanding of RNNs is sufficient or not. If you understand what I'm referring to then you shouldn't have any issues completing this project. The goal of the project for you is to both find the core issues with the existing code (I'd like to know what I did wrong) as well as provide a corrected version of the code which converges. Please start your bid description by typing out what the acronym, "RNN", stands for so I can weed out the auto-bids. Lastly, please provide a brief description of your experience with RNNs (include examples). I pay fast, I pay well, and I give bonuses for above and beyond service. Happy Bidding.