Thomas Su's Projects
This is code for an automated greenhouse that I created for a class at RPI. Although not all the code is written by me, I can certainly say that the vast majority is, and I mentored my peers in the process of writing the code.
This is some TCP and UDP work I did to explore the protocols
An example implementation of greedy knapsack versus dynamic knapsack
This repository attempts to determine the fair value of an equity-linked note (ELN).
This project implements a delta hedging strategy from a self-funding portfolio. As the delta of our portfolio changes, the portfolio is rebalanced such that the delta of the new portfolio is zero. Common random numbers (CRN) and the finite difference method is used to approximate delta. The repository pulls data from QuantMod to gather historical data and generates a market simulation using Heston's Stochastic Volatility Model with parameters generated from the gathered data. Our code works with European and Barrier options.
This code was made while I was a undergraduate researcher at CITE, Center for Infrastructure, Transportation, and the Environment at RPI, Rensselaer Polytechnic Institute. This code determines if a truck is moving or not based on GPS data from deliveries.
A quick exploration into natural language processing using kgrams
A really bad Othello AI. The AI isn't worth anything but maybe the board is? Sorry everything is in one huge file. This was made when I was a high schooler.
A command line shell for trading stocks using Robinhood
This is for an Intro to AI class I took at Rensselaer Polytechnic Institute. There are a lot of answers online already the homework for this class was taken from a open source UC Berkeley class so I don't think I'm helping anyone cheat... Anyways I think my code is nice.
In this repository we evaluate the performance of Stochastic Boosting and Traditional Boosting methods in two ways. The first is through evaluating the amount of data needed for each method to effectively generalize the classification problem. The second is effect of increasing the complexity of Weak Learner. How does a Weak Learner perform as it becomes for complex? Is it still able to generalize the classification problem in the same number of epochs?
This is a terminal simulator. It takes a text file as input, and runs the commands in the text file. Lots of good CS concepts like pipes, file descriptors, and threading.
This is a project that I did with The Center for Infrastructure, Transportation, and the Environment (CITE) at Rensselaer Polytechnic Institute (RPI). This is code creates multiple binary classifiers using various methods. Allowing a user to identify whether a given text is related to a disaster or not. Using a data that was provided by CITE to me, I was able to reach 95+% accuracy with certain methods. Some code is edited out to prodte
This is some code that I wrote for my Computer Hardware Design class. I would post more, but I'd hate to be a illegal answer key...