24F - DL Theory

EECE695D: Deep Learning Theory (Spring 2024) #

This semester, this course is:
(1) jointly served at POSTECH and Yonsei
(2) an online course
(3) taught in Korean

Team #

  • Instructor @ P. Jaeho Lee 이재호 ✉️
  • Instructor @ Y. Jy-yong Sohn 손지용 ✉️
  • TA @ P. Kyumin Kim 김규민 ✉️
  • TA @ Y. Chungpa Lee 이청파 ✉️
  • Contributor. Taesun Yeom 염태선 ✉️

Location & Time #

  • Class. Online at PLMS🔗 or LearnUs🔗.
    • Uploaded by 11:59PM on Wednesdays.
  • Office hrs. Use this sessions for Q&A.
    • @P. Wednesdays 2PM–4PM, Eng. Building#2 #323 (+ by appointment)
    • @Y. Mondays 1PM–3PM, Daewoo Hall #534

Schedule (tentative) #

  • W1. Overview, recap, and linear models
  • W2. Approx: Universal approximation with shallow nets
  • W3. Approx: Infinite-width and kernels
  • W4. Approx: Benefits of depth
  • W5. Optim: Convex optimization and generalizations
  • W6. Optim: SGD and flow-based analyses
  • W7. Optim: Implicit bias
  • W8. Gen: Concentration of measures, uniform convergence
  • W9. Gen: Rademacher complexities, covering numbers
  • W10. Gen: Chaining and VC dimensions
  • W11. Recent: Generalization
  • W12. Recent: Approximation
  • W13. Recent: Optimization
  • W14. Recent: Architectures
  • W15. Student Presentations - 1
  • W16. Student Presentations - 2

Further Readings #