Recent Advances in Deep Learning (EE807, Fall 2018)

Deep learning is a new area of machine learning research, which have demonstrated states-of-the-art performance on many artificial intelligence tasks, e.g., computer vision, speech recognition and natural language processing. In this lecture, I will cover some of recent advances (made mostly in the last 5 years) in this area.

Announcement

  • The maximum number of course registrations is around 40, and we cannot accept more now (do not send an email to Instructor or TA).

  • The deadline for notifyig whether you choose A or B to TA is Sep. 2 (see Lecture 0).

  • Two deadlines for sending your project reports to TA are Oct. 21 and Dec 16, which are all we need for your credit.

  • Office hours: Every Tuesday, 11am - 1pm, N1-914

Schedule

  • Lecture 0: Introduction to EE807

  • Lecture 1: Introduction to neural networks (DNN, CNN, RNN)

  • Lecture 2: Stochastic gradient descent (SGD)

  • Lecture 3: Regularization

  • Lecture 4: CNN architectures

  • Lecture 5: Generative deep models (RBM, DBM, SPN)

  • Lecture 6: RNN architectures

  • Lecture 7: Deep reinforcement learning

  • Lecture 8: Deep bayesian models

  • Lecture 9: Generative adversarial networks (GAN)

  • Lecture 10: Variational auto-encoders (VAE)

  • Lecture 11: Tranfer learning

  • Lecture 12: Domain translation

  • Lecture 13: Meta learning

  • Lecture 14: Zero/one/few-shot learning

  • Lecture 15: Novelty detection

  • Lecture 16: Adversarial examples

  • Lecture 17: Interpretable learning

  • Lecture 18: Network compression

  • Lecture 19: Application to vision I

  • Lecture 20: Application to vision II

  • Lecture 21: Application to language

  • Lecture 22: Application to vision + language (VQA)

  • Lecture 23: Application to speech

  • Lecture 24: Application to graphs (GNN, GCN)

Contact

Instructor: Jinwoo Shin, jinwoos@kaist.ac.kr
TA: Insu Han, insu.han@kaist.ac.kr