Type of Credit: Elective
Credit(s)
Number of Students
The course objectives are the in-depth discussions on issues regarding developing new learning algorithms of artificial Neural Networks on the infrastructure of PyTorch/TensorFlow and GPUs.
能力項目說明
The course objectives are the in-depth discussions on issues regarding developing new learning algorithms of artificial Neural Networks on the infrastructure of PyTorch/TensorFlow and GPUs. Particularly, we will study the PyTorch/TensorFlow framework that enables its programs performing parallel computations on GPU. Students will learn from the practice on developing new learning algorithms of artificial Neural Networks on the infrastructure of PyTorch/TensorFlow and GPUs. At the end of this course, students should gain: (1) the general knowledge on artificial neural networks methodologies, algorithms and implementation, and (2) the hands-on algorithm development experience on programs with PyTorch/TensorFlow and GPU.
| 教學週次Course Week | 彈性補充教學週次Flexible Supplemental Instruction Week | 彈性補充教學類別Flexible Supplemental Instruction Type |
|---|---|---|
|
週次 |
課程內容與指定閱讀 |
教學活動與課前、課後作業 |
學生每週學習投入時間 （含課堂教學時數） |
|
1 |
Hardware and Software (1) |
講義 / 閱讀 Stanford cs231n 2021 年版 Lecture 6 Hardware and Software、做作業 |
8 hrs/週
|
|
2 |
Hardware and Software (2) |
講義 / 閱讀 Stanford cs231n 2021 年版 Lecture 6 Hardware and Software、做作業 |
|
|
3 |
Training Neural Networks, Part 1 |
講義 / 閱讀 Stanford cs231n 2021 年版 Lecture 7 Training Neural Networks, Part 1、做作業 |
|
|
4 |
Training Neural Networks, Part 2 |
講義 / 閱讀 Stanford cs231n 2021 年版 Lecture 8 Training Neural Networks, Part 2、做作業 |
|
|
5 |
algorithm and BP |
講義、做作業 |
|
|
6 |
Learning goals and weight-tuning |
講義、做作業 |
|
|
7 |
Extra stopping criteria and an introduction of overfitting |
講義、做作業 |
|
|
8 |
Extra stopping criteria and an introduction of overfitting |
做作業 |
|
|
9 |
Overfitting -- regularizing |
講義、做作業 |
|
|
10 |
Overfitting – network-tuning |
講義、做作業 |
|
|
11 |
Learning dilemma -- isolating with ReLU |
講義、做作業 |
|
|
12 |
Learning dilemma -- analogizing with ReLU for binary-number inputs |
講義、做作業 |
|
|
13 |
Learning dilemma -- analogizing with ReLU for real-number inputs |
講義、做作業 |
|
|
14 |
The learning sequence and the initialization |
講義、做作業 |
|
|
15 |
New learning algorithms & AI application experiments |
講義、做作業 |
|
|
16 |
The validation experiment |
講義、做作業 |
|
|
17 |
Final project report |
|
|
|
18 |
Final project report |
|
Weekly meeting attendance (1% x 18) 18%
Meeting participation 36%
Group peer evaluation form 6%
Term project deliverable 40%
Total 100%
TBA
None