Machine learning is about computational methods that enable machines to learn concepts from experience. Many of the successful results of machine learning rely on learning with massive amounts of data labels. However, in many real problems we do not have enough labeled data, but instead have access to other forms of experience, such as structured knowledge, constraints, feedback signals from the environment, auxiliary models from related tasks, etc. This course focuses on those learning settings with few labels. This course is designed to give students a holistic understanding of related problems and methodologies (such as large language/multi-modal models, world models, self/weakly/un-supervised learning, transfer learning, meta-learning, reinforcement learning, adversarial learning, knowledge constrained learning, panoramic learning), different possible perspectives of formulating the same problems, the underlying connections between the diversity of algorithms, and open questions in the field. Students will read, present, and discuss papers, and complete course projects.

  • Lectures
    • Time: Tue/Thu 11:00am-12:20pm
    • Location: HDSI 123
    • Recordings: This classroom does not support podcast, so no recordings.
  • Discussion forum: Piazza (sign-up link)
  • HW and write-up submission: Gradescope (Entry Code: B3KGNG)
  • Contact: Students should ask all course-related questions on Piazza, where you will also find announcements. For personal matters or in emergencies, you can email the instructor.