機器學習原理及工業應用
Machine Learning and Industrial Application
| 節 | 週五 |
|---|---|
2 09:00–09:50 | 機器學習原理及工業應用 計中(3) 3 節連堂 |
3 10:10–11:00 | |
4 11:10–12:00 |
* 根據陽明交大上課時間表所列
This course is intended to equip the M.E. students with the followings 1. basic knowledge of machine learning principles 2. hand-on exercises to create a successful machine learning application 3. knowledge of Python and libraries 4. exposure to the real-world industrial (job) applications 5. presentation skills to promote his/her chosen ML subject(s) for an approval The objective of the course is to enable students to participate an industrial Machine Learning project at his/her work place. Some individual students may perhaps use Machine Learning methods (with Python) to conduct his/her own applications (industrial or otherwise). The content of the course includes the following 1. Python programming 2. Supervised Learning & methods 3. Unsupervised Learning & methods 4. Model Evaluation and Improvement* 5. Data Representation** 6. Complete hand-on experience (from data collection to result presentation) of his/her own chosen M.L. application *** (*) possibly taught with Unsupervised Learning (**) possibly taught with Supervised Learning (***) not on real industrial application (case study in class) unless his/her data collection/preparation consumes little time
(1) basic English reading and writing capability基本的英語讀寫能力 (2) programming experience on personal computers個人電腦上的編程經驗 (3) high school Math or Calculus 基本的數學能力或微積分 (4) access to a personal computer (e.g. computer lab) 有個人電腦可用
無備註
(1) PPT slides with hand-on coding, (2) some students will present M. L. homework with discussion in the class, (3) research on interested topic(s), collect data, apply M. L. knowledge and make some prediction in the final project, (4) present final project, (5) 2 助教, (6) 網站或圖書及資料庫given in the textbook and reference books, in addition to many U.S. university databases & websites (Google search)。
(subject to change given real situation): 6-7 Python/M.L. home works and 2 Quizzes (25%), 1 midterm or project (25%), 1 final project implementation (25%), final project presentation (25%)。
Python
(1) env. setup; (2) data types and operations; (3) flow control; functions; (4) file I/O
- 講授：
- 7
- 實作：
- 2
備註：2-3 HW to issue
Supervised Learning
(1) overfitting, underfitting, classification, convergence, (2) k-nearest neighbors algorithm, (3) linear model regression, ridge/lasso/ regression, linear model classification, logistic regression, (4) linear support vector machines (SVM,SVC), multiclass classification, naïve bayes classifiers, (5) decision trees, pruning, decision tree regression, random forests, gradient boosted regression trees, (6) neural networks, ANN with Keras.
- 講授：
- 12
- 實作：
- 3
備註：3-4 HW to issue, midterm
Unsupervised Learning
(1) clustering, PCA ( principal component analysis), (2) NMF (non-negative matrix factorization), (3) t-SNE, (4) k-means clustering, agglomerative clustering, dbscan
- 講授：
- 7
- 實作：
- 2
備註：1 HW to issue
Model evaluation
training set, validation set, test set, cross evaluation, 5-fold cross evaluation, grid search, etc.
- 講授：
- 2
備註：adjustable with supervised learning
Data representation
rescaling, categorical vs. continuous variable, binning, interaction & polynomial, transformation (std., robust std., min-max, normalizer)
- 講授：
- 3
備註：Adjustable with unsupervised learning
Student projects
Preview of topics, discussion, project presentation
- 示範：
- 3
- 實作：
- 3
備註：1-2 students per team; students be judges;\n
Industrial collaboration
TSMC panel discussion (subject to the availability of participating companies)
- 其他：
- 3
備註：TSMC projects; adjustable;
Introduction
Industrial applications (videos, etc.), syllabus, final project, etc.
- 講授：
- 2
| 週次 | 主題 |
|---|---|
| 第 1 週 | 1. overall review with syllabus - read "交通大學 2021S1 ML Syllabus_v1.pdf" 2. introduction of the class https://nycu.webex.com/nycu-tc/e.php?MTID=m49fc0a91f9ae75e689f81e3bf41c3266 5334 to enter 9/17 |
| 第 2 週 | 1. continue introduction to the class 2. teach Python & set up environment https://nycu.webex.com/nycu-tc/e.php?MTID=m7a850bdb52f08332f71847d80c7e0e70 5334 to enter 9/24 |
| 第 3 週 | 1. teach Python & set up environment 2. hand-on practice in class https://nycu.webex.com/nycu-tc/e.php?MTID=m2e9883d4db9b60b9c63fc335083a63bd 5334 to enter 10/1 |
| 第 4 週 | 1. teach Python 2. hand-on practice in class 3. issue HW2 10/8 |
| 第 5 週 | 1. teach Python 2. due HW1 3. issue HW2 10/15 |
| 第 6 週 | 1. Quiz#1 2. Python 3. Supervised Learning 10/22 |
| 第 7 週 | 1. Supervised Learning 2. KNN 3. Hand-on practice 4. issue HW3 10/29 |
| 第 8 週 | KNN practice/discussion Linear Regression Cross Evaluation 11/5 |
| 第 9 週 | midterm Ridge Regression issue HW4 11/12 |
| 第 10 週 | review Ridge/../HW4 Lasso Regression ElasticNet Logistic Regression 11/19 |
| 第 11 週 | multiclass classification naive bayes decision trees random forests gradient boosted regression trees 11/26 |
| 第 12 週 | kernelized support vector machines SVC( ) MLP (Neural Networks Industrial Example review students' final projects 12/3 |
| 第 13 週 | explain HW5, HW6 review students' final projects unsupervised learning - dimensional reduction: PCA, NMF, T-SNE 12/10 |
| 第 14 週 | Unsupervised Learning Clustering K-means, Agglomerative, DBSCAN, 3023 photos demo 12/17 |
| 第 15 週 | Final project presentations - 10 groups 12/24 |
| 第 16 週 | Final project presentations - 13 groups 12/24 |
(1) Introduction to Machine Learning with Python, by Andreas C. Müller & Sarah Guido (ISBN: 978-1449369415) (2) Hands-on Machine Learning with Scikit-Learn, Keras & TensorFlow, by Aurélien Géron (ISBN: 978-1492032649) (3) PYTHON 程式設計與數據分析/白文章 編著, 普林斯頓國際 (ISBN: 978-9869698917) (4) Practical Deep Learning 實用深度學習/謝哲光,鄭志宏,郭英勝 ,龔志銘,陳軒盈, CS滄海 (ISBN: 978-9863630722)
- 地點
- classroom or by appointment
- 時間
- Friday (after class) in Hsinchu, or NTUST in Taipei (by appointment)
- 聯絡方式
- (1) yearnhwang@nycu.edu.tw (2)0966-420208 (Taiwan cell) (3) LINE class group (by TA)