Headshot of Tavin Turner

Hi, I'm Tavin!

Live from snowy ColoradoπŸ”, I'm trying to make good things with technology. Wanna talk?πŸ“§

Right now I'm...

🦬 Studying CS at University of Colorado Boulder

πŸ‘€ Building super-secret new stuff (stick around)

πŸ§‹ Using computing for good with Blueprint Boulder

πŸ“» Skyrocketing my playlist count

Why I'm here

I'm here to educate the curious inside of me, and use that education to make the world a better place.

A rising technologist with conviction and rigor, from creative approaches in intelligent systems to accessible human-computer interaction.

With equity at the heart of scientific, social, and educational advancement, using technology to address the world's most pressing social and technological issues.

What I'm doing

Inner Product Methods in Neural Networks

April 2023 – Present

Foundationally, neural networks are not fully understood in many ways. The attempt to progress their comprehension opens the door to a rich ecosystem of modifying, reframing, and reconsidering the models according to our interests that has resulted in the dramatic and sudden diversity in neural learning models today. One assumption that has overwhelmingly remained true over their rapid growth, however, is use of the dot product to apply weights to a neuron’s input. We incise on the question: why the dot product? By framing inner products as measures of information explanation and correspondence, we can identify the neural network model as a non-linearized, repeated measure of explanation over the output of one layer and the weights of the next. We find that this complex transformation alters the spatial behaviors over models trained under otherwise identical conditions. This confirmation induces further exploration over how to engineer these spaces to specify that transformation, as well as, how to mitigate or harness asymptotic behavior.

Sentiment

June 2020 – June 2021

A sentiment-aware micro-journaling app unique in its two-dimensional sentiment analysis system capable of expressing emotion far beyond the capabilities of classifiers. Created in association with Computing for Good @ Colorado School of Mines.

TurboTransit, the Social Trip Planner

April 2023 – May 2023

End some emissions, crush that cash, and meet your mates! Uber meets Strava in this social-media-slash-trip-planner lifestyler! Whether it's that epic roadtrip or a dawn-breaking carpool, TurboTransit finds a local trip to your destination, then helps you screen your tripmates in mere minutes, so you can sail away easier than ever. Created in association with University of Colorado Boulder.

Game Advantage State Identification with Quantum Machine Learning

January 2021

We demonstrate the nearest centroid classification via implementations in 1) a fully classical model and 2) a model taking advantage of quantum distance estimation. Nearest centroid clustering was used to identify game state advantage in generated game boards, with comparable applications beyond the given application, from rapid medical classification to sentiment analysis. The quantum model is compared with the classical model to evaluate the efficacy of the quantum model in comparison to that of the classical model. Additionally, this comparison is gameified, pitting human, classical, and quantum classifications against each other in an interactive game.

Talk to me about...

βš›οΈ Quantum

🍡 Your favorite tea

🀝 Computing for good

πŸ—£οΈ Computational linguistics

πŸŽ“ Computer science education

🌐 Tuesday night's Wikipedia rabbithole

πŸ”¬ Your research, projects, company, or opportunities (I'm looking!)

πŸ’‘ Or whatever else you're working on