Honors Fall Poster Thesis Symposium

Fall Poster Symposium

1:30 p.m.—Nathaniel Patterson (Zoom)
Finding an Embedding for Music Auto-Complete: An LSTM Approach

Recent success in sequence generation has been seen in the problem of Polyphonic Music Generation. This study analyzes the effectiveness of two embedding strategies: notes as a string and notes as objects using a Long-Short Term Memory Recurrent Neural Network (LSTM-RNN) for music auto-completion when trained on the corpus of Erik Satie. This project seeks to introduce Music Autocomplete as a new problem, while adding to the body of knowledge on how Neural Networks process sequential data and how different data embeddings improve performance. This project also adds to the subfield of the intersection of artistry and Artificial Intelligence.

2 p.m.—Lloyd Martinez (in-person in NH 143)
The Effective Predictors of the SPX Index

We use a mathematical model based on a multiple linear regression to investigate the influence of different variables (Option Difference between Puts and Calls, Volatility Index, Interest Rate on Required Reserves, 10-Year Treasury Minus 2-Year Treasury Spread, ICE BofA Option-Adjusted Spreads (OASs), Gold Currency, and Trade Weighted United States Dollar Index) to explain the variance in closing price of the SPX. In our regressions, we find statistically significant relationships linking our independent variables to the closing price of the SPX on a daily, weekly and monthly basis. The forecast error evaluations of our models suggest that investors can accurately predict financial markets to a certain extent.

*Masks required. A limited number of guests would be most welcome to attend.