New Course Focuses on Data-Centric AI
TL;DR. A new course is being offered on Data-Centric AI, an emerging field that studies techniques to improve datasets for better machine learning model performance. - The course teaches practical algorithms to fix common issues in ML data and construct better datasets. - It covers topics like label errors, class imbalance, distribution shift, and data curation for LLMs. - The material emphasizes real-world ML applications and includes hands-on programming exercises.
- Introduction to Data-Centric AI is the first-ever course dedicated to the subject.
- It teaches systematic engineering disciplines for improving datasets in practical ML applications.
- The curriculum addresses issues such as label errors, class imbalance, and data curation for LLMs, with practical labs.
Sources
- Introduction to Data-Centric AI — dcai.csail.mit.edu