Cornell-Citi Financial Data Science Webinar with Nicholas Venuti (Morgan Stanley)
For the final Cornell-Citi Financial Data Science Webinar of the semester, we are happy to introduce Nicholas Venuti of Morgan Stanley on Tuesday, 5/11. His presentation, "Advances in Sequential Deep Learning," will be held live via Zoom from 5:00pm to 6:00pm EDT.
This webinar is free and open to all guests. Registration is required (Registration Link). You will receive the webinar link and dial-in info upon registration (the confirmation email will come from firstname.lastname@example.org).
Abstract: Sequential data serves as the basis for many real-world applications such as machine translation, voice-to-text conversion, and motion tracking. As the order and context of past datapoints are needed for future prediction, these datasets must be modelled temporally either by constructing features or utilizing recursive models.
Many state-of-the-art solutions include recurrent neural networks (RNNs), as these methods are able to exploit both techniques by capturing the time-based nature of these systems while leveraging the expressiveness of deep learning architectures. While RNN variants such as gated recurrent units (GRUs) and long-short term memory (LSTM) networks have dominated sequential deep learning, recent studies have found that temporal convolutional networks (TCNs) can match or exceed these networks in prediction performance while greatly reducing the training time of the models.
In this talk we will provide a general overview of four common architectures: vanilla RNNs, GRUs, LSTMs, and TCNs. We will compare the model stability, memory requirements, and training times of each. Lastly, we will review the performance of these architectures on a variety of benchmark image, text, and audio datasets.
1) Nicholas Venuti's Presentation
3) "Lightning Talk" featuring CFEM Alumnus Vineel Yellapantula
Speaker Bio: Nicholas Venuti is a Machine Learning Research Scientist in Morgan Stanley's Machine Learning Center of Excellence, whose main research focus is deep learning architectures for time-series predictions. After obtaining his Bachelor of Science in Biomolecular Chemical Engineering at North Carolina State University, Nicholas began his career working in data analytics at an environmental consultancy. Afterwards, he obtained his Masters of Data Science at the University of Virginia, where his thesis studied using NLP to identify semantic shifts in religious and political texts as an early indicator for extremist views.
"Lightning Talk" Info: CFEM alumnus Vineel Yellapantula will discuss his summer project at AbleMarkets under Prof. Irene Aldridge, “Quantifying Sentiment in SEC Filings.” By utilizing Natural Language Processing techniques and the BERT model, he explored how text present in the MD&A section of 10-K and 10-Q filings affect the performance of the stock. He also tested the efficacy of multiple factors derived from these texts using a long-short market-neutral trading strategy.
Vineel Yellapantula (MFE Cornell ’20, MSc Mathematics BITS Pilani '18) is a Decision Analytics Associate at ZS Associates.