大型語言模型
Large Language Models
| 節 | 週五 |
|---|---|
5 13:20–14:10 | 大型語言模型 ED103（光復） 3 節連堂 |
6 14:20–15:10 | |
7 15:30–16:20 |
* 根據陽明交大上課時間表所列
Large language models (LLMs) offers a unique opportunity to peek behind the curtains of the groundbreaking advancements in generative artificial intelligence (GAI). By delving into these developments, you will understand how language models are evolved and exploited to deploy over various domains. This course will address how LLMs work under the hood, tearing the lid off the GAI black box. LLMs provide a platform to explore how machines can comprehend, generate, and manipulate language in ways that mimic human cognition. This course focuses on the fundamentals in computation theories, architectures and practices for LLMs, and highlights the academic and industrial advances in the extended models and applications.
Calculus, Linear Algebra, Probability and Statistics
無備註
Teaching materials, codes and datasets will be provided. Teacher assistants will be available at PM19:00-20:00 in week days. Appointments are required. You are encouraged to use online discussion function in E3. TAs will promptly reply your questions.
Temporary Policy: Homework (or Task Competition) (60%), Final Project (40%), Class Attendance (+10%)
-
Machine Learning Basics
-
Neural Network Models
-
Optimization for Deep Models
-
Recurrent Network and Attention
-
Generative Models, Transformer, BERT, GPT
-
State-Space Model and Diffusion Model
-
DaVinci GAI Platform and Applications
-
LLM Model Trends and AI Computing Architecture for LLM
| 週次 | 主題 |
|---|---|
| 第 1 週 | Feedforward and Convolutional Neural Networks 2025-09-05(五) 時數：[2025-09-05]簡仁宗(3.00) |
| 第 2 週 | Regularization for Optimization in Deep Learning 2025-09-12(五) 時數：[2025-09-12]簡仁宗(3.00) |
| 第 3 週 | Recurrent Neural Network and Sequential Learning 2025-09-19(五) 時數：[2025-09-19]簡仁宗(3.00) |
| 第 4 週 | N-Gram Language Models and Topic-Based Language Models (Grouping) 2025-09-26(五) 時數：[2025-09-26]簡仁宗(3.00) |
| 第 5 週 | RNN Language Models and Language Understanding (1st Homework) 2025-10-03(五) 時數：[2025-10-03]簡仁宗(3.00) |
| 第 6 週 | Attention Networks and Transformers 2025-10-10(五) 時數：[2025-10-10]簡仁宗(3.00) |
| 第 7 週 | State-Space Language Models (Project Proposal) 2025-10-17(五) 時數：[2025-10-17]簡仁宗(3.00) |
| 第 8 週 | BERT Encoder and GPT Decoder 2025-10-24(五) 時數：[2025-10-24]簡仁宗(3.00) |
| 第 9 週 | Retrieval, Augmentation and Generation 2025-10-31(五) 時數：[2025-10-31]簡仁宗(3.00) |
| 第 10 週 | Generation with Prompting Strategies (2nd Homework) 2025-11-07(五) 時數：[2025-11-07]簡仁宗(3.00) |
| 第 11 週 | LLMs with GPT and LLaMA 2025-11-14(五) 時數：[2025-11-14]簡仁宗(3.00) |
| 第 12 週 | LLM Model Trends and Generative AI 2025-11-21(五) 時數：[2025-11-21]簡仁宗(3.00) |
| 第 13 週 | Final Presentation 2025-11-28(五) 時數：[2025-11-28]簡仁宗(2.00) 梁伯嵩(1.00) |
| 第 14 週 | AI Computing Architecture for LLM 2025-12-05(五) 時數：[2025-12-05]梁伯嵩(3.00) |
| 第 15 週 | Final Presentation 2025-12-12(五) 時數：[2025-12-12]簡仁宗(2.00) 梁伯嵩(1.00) |
| 第 16 週 | DaVinci GAI Platform and Applications 2025-12-19(五) 時數：[2025-12-19]梁伯嵩(3.00) |
1. Lecture Notes and Slides 2. I. Goodfellow and Y. Bengio and A. Courville, Deep Learning, The MIT Press, 2016. 3. S. Watanabe and J.-T. Chien, Bayesian Speech and Language Processing, Cambridge University Press, 2015.
- 地點
- ED708
- 時間
- PM18:00-18:30 on Monday. Appointments are required.
- 聯絡方式
- jtchien@nycu.edu.tw