Taboola was excited to sponsor the recent Hacktech Hackathon at the California Institute of Technology in Pasadena, CA. Over 500 students from around the country came together for the event, an intense 36-hour competition where teams of up to four students hacked away through the night, submitting nearly 100 tech projects for prizes topping $32,000.
Each team came to the competition with a rough project idea and worked to build from it a functioning prototype. With high-value credits from sponsors and cash prizes at stake, teams realized and refined ideas they had previously only theorized.
Taboola’s sponsorship focused on the best use of machine learning, big data, or data processing; nearly a quarter of all projects applied for our prize. One team created a way to provide secure encryption that ensures recognition of a user’s precise handwriting style (using a neural network and Synaptic touch sensor).
Another team combined a Microsoft Kinect with machine learning to understand and analyze a person’s posture as they worked. We also saw a fun project which utilized a muse headband (or EEG brain sensor) and machine learning to create a new kind of lie detector test.
Given each project’s level of sophistication, and the short time frame allotted for producing a workable prototype, it was easy to forget that the participants were still students, a fact only hinted at by the Hackathon’s intermittent donut runs and game breaks.
Ultimately, the award for “Best Use of Machine Learning” went to the SkinDeep team whose web application uses predictive technology to classify skin lesions as cancerous or benign. The prototype allows users to submit a picture of their skin abnormality through a web application interface; then, a set of algorithms extract relevant features from the image and return a classification.
The team’s technical process to generate this classification proved very impressive. User images are sent to a server for feature extraction, where three self-developed image processing algorithms collect up to nine relevant features, including eccentricity, symmetry, color variation, and more.
The image is then sent to Microsoft Azure servers and analyzed by a Two-Class Support Vector Machine, an algorithm that has been trained on a public index of thousands of relevant dermatological images.
In return for their winning prototype, each of the SkinDeep team members were awarded with a new drone. The project was created by three computer science majors: Evan Johnson, a freshman at Northeastern University (left); Patrick Pan, a sophomore at Harvard (center); and Ashwin Varma, a sophomore at Rice University (right).
Hackathons have become a popular event on college campuses in recent years, attracting competitive students who are excited to show off their tech skills and develop new ones. These events call on hackers to use libraries and interfaces (provided by sponsors or available online publicly) to build a solution to a problem in a limited amount of time.
Common interfaces include Google Maps, texting platforms, cloud processing, and hardware sensors. Most participants study computer science at various colleges and high schools around the country (though college students account for the majority of participants).
Sponsors often have team members on-site who help mentor hackers as they build their idea into a working project. Hackathons are selective in their admission process, seeking out teams whose ideas exhibit innovation and dedication, and the events are an opportunity for tech leaders as well as students.
Representatives from leading tech companies dedicate time towards connecting with students and encouraging high-quality candidates to pursue engineering as a career.
In the coming months, Taboola will be sponsoring LA Hacks, one of the largest hackathons in the country, at UCLA. If you’ll be attending the event, don’t forget to stop by our booth to say “hi” while you’re there!
(Picture credits: Wolfram, Gemma Busoni)