#### Advanced Deep Learning (Adv DL) – CISC889013 – Spring 2024 – ISE Lab 205

**Prerequisite**: Formally, CISC681; practically, any 400+-level courses in AI/ML

**Course Description:** This course will cover advanced deep-learning topics. We will start by very briefly covering the basics of neural networks and deep learning. We will cover multilayer perceptrons (MLPs), convolutional neural network (CNNs), and recurrent neural network (RNNs).

After this, out of many possible topics, the course will cover foundation models (transformers and large language models), latent variable models (variational autoencoders and diffusion models), generative adversarial networks, graph neural networks, and deep reinforcement learning. In each topic, we will cover the core theoretical concepts and the most prominent variations of implementing each type of model.

The class will include lecture sessions (covering the theory and variations), followed by paper reading sessions (covering papers picked by the students).

##### Topics [Tentative]

Date | Topic | Key models covered |
---|---|---|

2/6 | Logistics, DL Basics | MLP |

2/8 | DL Basics | CNN |

2/13 | DL Basics | RNN |

2/15 | LLMs Theory | Transformer |

2/20 | LLMs Theory | GPT, BERT, CLIP |

2/22 | LLMs - Paper review | |

2/27 | LLMs - Paper review | |

2/29 | Variational models - Theory | |

3/5 | Variational models - Theory | Beta VAE, VQ-VAE |

3/7 | Variational models - Paper review | |

3/12 | Variational models - Paper review | |

3/14 | Generative Adversarial Networks - Theory | Wasserstein GAN, InfoGAN, CycleGAN, StyleGAN, BigGAN |

3/19 | Diffusion Models - Theory | DDPM |

3/21 | Diffusion Models - Theory | DALL-E |

3/26 | No Class - Spring Break | |

3/28 | No Class - Spring Break | |

4/2 | Generative Models - Paper review | |

4/4 | Generative Models - Paper review | |

4/9 | Graph Neural Networks - Theory | |

4/11 | Graph Neural Networks - Theory | GCN, GAT |

4/16 | Graph Neural Networks - Paper review | |

4/18 | Graph Neural Networks - Paper review | |

4/23 | Deep Reinforcement Learning - Theory | |

4/25 | Deep Reinforcement Learning - Theory | DQN, PPO |

4/30 | Deep Reinforcement Learning - Paper Review | |

5/2 | Deep Reinforcement Learning - Paper Review | |

5/7 | Project Presentation | |

5/9 | Project Presentation | |

5/14 | Project Presentation | |

5/16 | Project Presentation |

##### Learning Objectives

- For the five major deep learning topics covered in class, be familiar with the following aspects:
- Theoretical (math) basis
- Algorithmic implementation and considerations
- Major applications and success stories
- Current SOTA methods
- Any possible societal impacts and implications
- Future ways of extension and research

##### Papers

##### Projects

##### Grading

Category | Percentage | Notes | |
---|---|---|---|

Reading assignments | 40 | 2 papers of your choice per topic | |

Project | Code (+ short reprot) | 15 | |

Presentation | 10 | ||

Participation | 20 | ||

Knowledge retention mini-exam | 5 | 2 PM, May 19. Same Class | |

Peer-review | 10 | For grading classmates |

Grade | Interval | Grade | Interval | Grade | Interval |
---|---|---|---|---|---|

A | 90 and over | B- | 70 to 75 | D+ | 50 to 55 |

A- | 85 to 90 | C+ | 65 to 70 | D | 45 to 50 |

B+ | 80 to 85 | C | 60 to 65 | D- | 40 to 45 |

B | 75 to 80 | C- | 55 to 60 | F | Below 40 |

##### Ed Discussion

We will use Ed for ALL of the class communications. The system is highly catered to getting you to help fast and efficiently from classmates, the TA, and myself. Rather than emailing questions to the teaching staff, I encourage you to post your questions on Ed. Check for existing questions, and consider answering your classmates’ questions.