How to do more with less: an Active Learning perspective
One common scenario for developers training Machine Learning models is the lack of available annotated data. Models require vast amounts of data to perform well and be robust. Active Learning promises a solution to this problem by maximizing a model's performance while reducing the size of the training corpus. In this presentation, you will learn what Active Learning is and how to use it in your projects to reduce costs and time spent on data annotation. I will discuss my own work in the field, including a novel technique inspired by a natural algorithm (hint: it involves intelligent ants).
Andrei-Robert Alexandrescu
Computer Science master's student
- Babes Bolyai University
Andrei is a final-year Computer Science master's student specializing in Applied Computational Intelligence. His primary interests are in Computer Vision for autonomous driving systems and Active Learning for reducing data annotation costs. He has published several articles on these topics and presented at international conferences such as VISAPP and YTIC, with an upcoming presentation at KES 2024.
Andrei's professional experience is diverse, encompassing research with Bosch on autonomous trains and software engineering/developer internships at Google, Amazon, and RebelDot. He is also an active community member, serving as a co-lead for the Google Developer Student Clubs @ BabeČ™-Bolyai University. In this role, Andrei organizes and hosts technical events for students on various Google technologies. Additionally, he has delivered numerous presentations on both technical and non-technical subjects at various events.