Statistical Topics in Modern Machine Learning (FGV EMAP, Rio)

Schedule

From April, 15th to June, 15th

• Tuesdays 9:20am - 11am, Auditório 318
• Tuesdays 4:20pm - 6pm, Sala 1014
• Fridays 2:20pm - 4pm, Sala 1014

Lecture notes

Written in collaboration with Claire Boyer, Ismaël Castillo and Étienne Roquain.

1. Starter on neural networks
2. Approximation properties of neural networks
3. Complexity of neural networks
4. Regression with neural networks, ERM and minimax lower bounds
5. Generative adversarial networks
6. Diffusion models
7. Confidentiality and privacy-preserving inference
8. Conformal inference
9. Reproducing kernel Hilbert spaces, Neural tangent kernel
10. Double descent and benign overfitting