Francesco Orabona wins CAREER award for work on machine learning

Assistant Professor Francesco Orabona (ECE, SE, CS) won the National Science Foundation CAREER award for his work on new, more automated, machine learning algorithms.

Machine learning has begun to take over our digital lives. It can be found in automatic text suggestions in email, recommendations for the next show to binge-watch on Netflix, and even self-driving cars. Through machine learning, a computer can use algorithms to teach itself how to imitate an input or achieve a goal in the most efficient and accurate way possible, but right now those programs still need parameters that are adjusted by a real person in order to function. Orabona wants to fix this final inefficiency, by finding an algorithm that allows machine learning to function completely autonomously.

ECE Assistant Professor Francesco Orabona

“We had these imperfect machine learning algorithms for so long that people are not even aware that something better exists,” Orabona said. “So, a big chunk of my work is in disseminating these ideas on parameter-free machine learning algorithms.”

Orabona’s passion for achieving perfectly-automated machine learning began while he was getting his PhD in robotics. When he and his classmates had live demos of their robot doing visual and manipulation tasks, he was often the one doing the tedious task of manually setting the parameters for the demonstration.

“I used to go into the lab one hour before and find by trial and error the perfect settings of all the parameters of the algorithms for the current light conditions, object to manipulate, etc. It was not a fun job,” Orabona said. “After my PhD, I moved to the area of machine learning but that experience stayed with me. So, I devoted myself to making these kinds of algorithms truly automatic.”

He sees his CAREER award as “essential step in my career” and is grateful for the recognition from the community for his work so far, but still, “there is so much to do!” Orabona updates people on the progress of parameter-free machine-learning algorithms on his blog, hoping to get more people interested in the research.

“The more people work on it,” Orabona said, “the faster we will have better algorithms!”

Professor Francesco Orabona got his PhD from the University of Genoa in 2007 and manages the Optimization and Machine Learning Lab at Boston University. He is a 2017 Google Research Award winner and won Best Paper at the International Conference on Image Analysis and Processing in September 2015.