Loss Landscape and Performance in Deep Learning by Stefano Spigler

Loss Landscape and Performance in Deep Learning by Stefano Spigler

International Centre for Theoretical Sciences via YouTube Direct link

Set-up: Dataset

4 of 21

4 of 21

Set-up: Dataset

Class Central Classrooms beta

YouTube videos curated by Class Central.

Classroom Contents

Loss Landscape and Performance in Deep Learning by Stefano Spigler

Automatically move to the next video in the Classroom when playback concludes

  1. 1 Loss Landscape and Performance in Deep Learning
  2. 2 Supervised Deep Learning
  3. 3 Set-up: Architecture
  4. 4 Set-up: Dataset
  5. 5 Learning
  6. 6 Learning dynamics = descent in loss landscape
  7. 7 Analogy with granular matter: Jamming
  8. 8 Theoretical results: Phase diagram
  9. 9 Empirical tests: MNIST parity
  10. 10 Landscape curvature
  11. 11 Flat directions
  12. 12 Outline
  13. 13 Overfitting?
  14. 14 Ensemble average
  15. 15 Fluctuations increase error
  16. 16 Scaling argument!
  17. 17 Infinitely-wide networks: Initialization
  18. 18 Infinitely-wide networks: Learning
  19. 19 Neural Tangent Kernel
  20. 20 Finite N asymptotics?
  21. 21 Conclusion

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.