Have a hard drive full of sequential data like audio, text, images or even DNA, but you can’t figure out how to make the most of it? Have an innovative research solution or business objective in mind, but you’re stuck in a quagmire of summarizing data into a few condensed features? Need broad solutions that cover a number of tasks—for example, generating conversational responses based on text or audio data?
This summer, the Data Science Institute will host “AI Summer,” a four-week interactive, hands-on learning experience with live coding where anyone with an interest in advanced data science techniques can learn to use, train and share data models using any research or project data.
DSI AI Summer is open to graduate students, postdocs, researchers, faculty, undergraduates, staff and the Nashville community. The sessions are free, and registration is required.
The courses will be May 10–June 3, on Mondays, Wednesdays and Fridays from 9 a.m. to noon CT. The first week of programming is an optional hands-on learning opportunity for attendees new to programming or to Python—the high-level, general-purpose programming language that will be used throughout the sessions.
Sessions will focus on the powerful and flexible models called transformers. With the infusion of context, domain understanding and standardized feature spaces, transformers harness the richness of complex datasets for problem-solving using a single pre-trained model, sometimes with no additional training on your data (“zero-shot” solutions). From text to images and audio (“textless natural language processing”), transformers can identify objects in images, identify concepts in spoken conversations, ask questions about tables of data and more.
DSI AI Summer participants will walk away with a high-level understanding of transformers, their building blocks and their relationship to other deep learning architectures and guided practice in training and using transformer models with multimodal data. DSI staff data scientists will hold office hours to discuss individual research projects, and participants can build new partnerships with faculty and students in other disciplines for collaboration, undergraduate research, immersion and data sharing.