본 수업에서는 최근 기계학습 및 심층신경망 학습의 한 분야로 대두되고 있는 깊은 생성 모델(deep generative models, DGM)들을 다루고 있다. DGM은 데이터를 통해 새로운 가상의 데이터를 생성하는 것을 목표로 하는 모델로 본 수업에서는 기초적인 generative adversarial network(GAN)과 variational auto encoder(VAE) 모델들부터 최근 제안된 flow, energy-based, diffusion 모델들을 함께 다룰 예정이다.

Please check the following webpage for the detailed schedule and others.

https://sites.google.com/view/dgm-2022-postech-kaist/

You must take the machine learning and deep learning courses before taking this course.

Homework: 20%
Final Exam: 20%
Fianl Project: 60%

Week1: Course OT / Bayesian Network
Week2: Expectation Maximization / Variational Inference
Week3: Variational Inference
Week4: VAE
Week5: VAE / Introduction to Potential Tasks by Deep Generative Model
Week6: GAN
Week7: Short Intro to Differential Equations and Integrals / GAN
Week8: Project presentation week
Week9: GAN / Flow
Week10: Flow
Week11: Flow / Neural ODE
Week12: Physics and ML / Short Intro to Stochastic Differential Equations and Integrals
Week13: Hierarchical and Quantized VAE / Diffusion Models
Week14: Diffusion Models
Week15: Final project presentation
Week16: Final exam

hybrid lecture (online + offline)
본 강의는 카이스트와 공동 교과 수업으로 주차별로 포스텍 및 카이스트에서 현장 이론강의를 제공하며 이를 온라인으로 생중계 할 예정.
포스텍과 카이스트 수업시간 차이로 인하여 수업시간은 오후 2시 15분에서 3시 30까지 진행될 예정.

Course homepage: https://sites.google.com/view/dgm-2022-postech-kaist/

Grading criteria are shared. Gradings are separated between POSTECH and KAIST.
i.e.) POSTECH students will be only evaluated and letter-graded among POSTECH peers.

Homework : 20%
Final Exam : 20%
Final Project: 60%

Subject to change bu the students’ performance and participation