ESE PhD Candidate Liangzu Peng defended his thesis on Ideal Continual Learning: Theory and Practice on Tuesday, August 25th. Advised by Professor Rene Vidal, Peng’s research contributes to the continual learning community by developing some ideas for understanding how models can learn new tasks without forgetting previously acquired knowledge.
“My work identifies the information from past tasks that must be preserved and provides efficient algorithms and theoretical guarantees for doing so,” Peng explains. “This helps bridge the gap between the theory and practical design of adaptive learning systems that can improve continuously.”
Throughout his doctorate, Peng experienced challenges with work-life balance, particularly when it came to caring for his child. “Childcare was time-consuming and often fragmented my day, making it difficult to sustain the deep concentration that research requires,” Peng overcame the above challenges by becoming more intentional with time management. “My wife and I also coordinated our childcare responsibilities so she could care for our child for a few hours at a time, allowing me to focus on research during those periods.”
Peng expressed gratitude to several people for their support throughout his research journey. First and foremost, Peng thanks Professor Rene Vidal. Peng’s gratitude extends to family members such as his grandmother, wife, friends, lab members as well as staff from both the ESE Department and IDEAS Center. In addition, Peng thanks organizations that support his research such as the Research Council of Norway, the National Science Foundation, the Simons Foundation, the SEAS Dean’s Fellowship, and Penn AI Fellowship.
While Peng has countless memorable moments of his time at Penn, walking home from the Van Pelt Library through Locust Walk to his home represents one of the most vivid. “The trees lining the path made the walk feel peaceful and refreshing, especially on sunny days, when sunlight filtered through the leaves and cast shifting patterns of light and shadow along the walkway.”
Future plans for Peng include joining a stealth AI startup in London. There, Peng will “conduct research at the frontier of scientific discovery. I expect the job to be fun and challenging. I promise I’ll work out more and stay healthy.”
When Peng’s not working, he enjoys “ having fun with my 18-month-old, cooking, and chatting with chatbots to better understand life, meaning, and the world.”
Learn more about Peng’s work here