Omkar Mehta's repositories
pyspark-tutorial
Learn PySpark for all future projects.
SSFwithNODE
Sub-seasonal Forecasting with Neural ODE.
tensorflow-tutorial
# TensorFlow Tutorial ## © [Omkar Mehta](omehta2@illinois.edu) ## ### Industrial and Enterprise Systems Engineering, The Grainger College of Engineering, UIUC ### <hr style="border:2px solid blue"> </hr> Until now, I've always used numpy to build neural networks. Now I will step through a deep learning framework that will allow myself to build neural networks more easily. Machine learning frameworks like TensorFlow, PaddlePaddle, Torch, Caffe, Keras, and many others can speed up machine learning development significantly. All of these frameworks also have a lot of documentation, which I feel that it's fun to read. I will learn to do the following in TensorFlow: - Initialize variables - Start my own session - Train algorithms - Implement a Neural Network Programing frameworks can not only shorten my coding time, but sometimes also perform optimizations that speed up my code.
Classification-on-textual-data
Main purpose of this project is to understand different methods of classifying textual data. I've used "20 newsgroup" dataset. It is a collection of approximately 20,000 newsgroup documents, partitioned (nearly) evenly across 20 different newsgroups, each corresponding to a different topic.
entropy-and-persistent-homology
To begin with, I am interested in ways to think about diversity and geography. One example might be variants of Covid-19, another might be trees ([https://senseable.mit.edu/papers/pdf/20210325_Galle-etal_MappingDiversity_UFUG.pdf](https://senseable.mit.edu/papers/pdf/20210325_Galle-etal_MappingDiversity_UFUG.pdf)). I would like to see if wildlife data, as in [https://www.inaturalist.org/](https://www.inaturalist.org/), can support the same type of analysis. Compared to trees, wildlife is perhaps a bit more interesting since some wildlife is carnivorous. I am interested in combining/comparing (in a way I haven't yet figured out) entropy (see the above paper) and persistent homology (see [https://projecteuclid.org/journals/algebraic-and-geometric-topology/volume-7/issue-1/Coverage-in-sensor-networks-via-persistent-homology/10.2140/agt.2007.7.339.full](https://projecteuclid.org/journals/algebraic-and-geometric-topology/volume-7/issue-1/Coverage-in-sensor-networks-via-persistent-homology/10.2140/agt.2007.7.339.full)). See also [https://www.youtube.com/watch?v=8nUBqawu41k](https://www.youtube.com/watch?v=8nUBqawu41k). Task: can you download data from [https://www.inaturalist.org/](https://www.inaturalist.org/) so that we can analyze it? We want: date of observation latitude/longitude of observation what is being observed (we in fact want to use some sort of tree of life classification, as in [https://a-z-animals.com/reference/animal-classification/](https://a-z-animals.com/reference/animal-classification/)) This problem is admittedly still in its infancy. The immediate win for you would be learning to poke around datasets. An intermediate win would be visualization and understanding some new tools. If things work out, I think that this could be an unexpectedly novel combination of tools applied to an interesting problem.
optimization-methods
Python implementation of the optimization methods:
sign-language-classification
Build Convolutional Neural Networks from scratch
stanford-cs231n-assignments-2020
This repository contains my solutions to the assignments for Stanford's CS231n "Convolutional Neural Networks for Visual Recognition" (Spring 2020).
style_transfer_webapp
https://web-app-style-transfer-hpn4y2dvda-uc.a.run.app/