Week 1: Working with Python

Feb 18, 2019

Last summer, I had the opportunity to conduct research at the University of Iowa, and having thoroughly enjoyed that experience, I’ll be continuing working with my advisor for my senior project. Over five weeks last summer, I uncovered how black-box neural networks make decisions and now, during the third trimester of senior year, I’m taking the next step: increasing predictive accuracy via a hybrid model, while maintaining transparency in the decision-making process.

This first week, I jumped headfirst into coding. After a meeting with my external advisor, Dr. Lin, and my internal advisor, Ms. Bhattacharya, I decided to spend the majority of my week learning Python and translating my code from the summer (written in R) to Python. Setting up my Python IDE, learning basic commands, working with terminal, and finding the Python equivalent to each line of my R code was a lot more tedious than I expected. I ran into numerous problems regarding importing/installing packages, duplicating the neural network from the summer, getting the dataset I needed, and more, but with the help of my advisors, StackOverflow, Github, and Google, I was able to find solutions to most of them. This week set my expectations for this project straight: it’ll be a great learning experience, but will definitely have some hurdles along the way.

In addition, I read my first paper this week: Chapter 10: Linear Neural Networks. All my previous interactions with neural networks have been with black-box models, so learning about the linear models was an important step in the process of building a hybrid neural network model.

Thanks for reading my first post, and follow along for weekly updates on my project!

4 Replies to “Week 1: Working with Python”

  1. Vidur G. says:

    This project sounds highly interesting! Neural networks are the future of computing and I am excited to see what happens next!

  2. Neural networks sure are a black box in many ways. What is unique about hybrid neural networks specifically?

    1. Arshia Singhani says:

      I hope to strike a balance between interpretability/explainability of neural networks (which comes with linear models) and accuracy (which is better with black-box models)!

  3. Ivana B says:

    It is good to realize early this is not going to be a smooth ride. You may or may not reach your desired goal, but the lessons you learn along the way will be valuble.

Leave a Reply

Your email address will not be published.