Caltech Frosh Wins OTT—Intellectual Ventures Invention Competition

Caltech freshman Vighnesh Leonardo Shiv loves music and computer science. That pair of passions led him to envision a machine-learning tool that has earned him the $25,000 top prize in this year's Intellectual Ventures–Caltech Invention Competition.

The contest is based on the idea that today's young inventors are tomorrow's technological and entrepreneurial leaders. In this, its second year, the competition encouraged undergraduate students to create a product or process, apply technology in a new way, or otherwise demonstrate remarkable inventiveness. Graduate students could participate as supporting team members. The competition was sponsored by Caltech's Office of Technology Transfer and by Intellectual Ventures, a privately held company based in Bellevue, Washington, which focuses on the business of invention.

"Sponsoring the competition at Caltech again this year was an honor for Intellectual Ventures," says Robert Lee, a portfolio group technology analyst with the company and a Caltech alumnus. "We believe ideas have value and strive to bring more inventions into the world by making the act of inventing both scientifically interesting and commercially viable. We were impressed by the high quality and scientific breadth of student submissions this year."

Shiv's proposal, titled "Continuous-Weight Neural Networks," describes a tool that could more effectively analyze and mine data for useful information. In computing, neural networks are tools that analyze data with given classifications and generalize features within that data so that they can make predictions about how to classify unlabeled data. Shiv's award-winning idea is an algorithm that integrates two neural networks, nesting them in such a way that the algorithm can analyze smaller sets of features while retaining needed information, leading to better learning.

Originally from Portland, Oregon, Shiv played in bands growing up and sings and plays the guitar in a hard-rock band at Caltech. In high school, when thinking about a research project to undertake, he was intrigued by the problem of music transcription, the translation of heard music into notes on a page. "The idea for the project really came out of my love for music," he says. "No one else in the bands I was in could really play by ear, so I always had to transcribe all the music for everyone. So I thought, 'Let's get a computer to do it instead.'"

Admittedly, he says, his invention is only vaguely related to that original topic, but it could have applications in everything from speech recognition and music processing to computer vision and augmented reality. 

"As a computer scientist, I really like inventing algorithms," he says. "Having an outlet for that in this contest was really, really cool."

Second place in the competition and $15,000 went to senior Samson Chen and sophomore Chris Varnerin, who suggested a new approach to touchscreen technology. Third place and $10,000 went to senior Alexander Hu and graduate students Steven Bowers, Kaushik Dasgupta, and Kaushik Sengupta. Drawing inspiration from the biological world and the ability of living things to heal wounds, they designed a circuit with its own fully integrated system for self-healing.

For more information on the competition, click here.