2018-2 Artificial Intelligence
Class Info:
Time: Class1 (23550): Tue 10:00-13:00, Class2 (23551): Thr 13:00-16:00
Location: Cluster Bd (학연산클러스터), Room 509
Textbook: Python Machine Learning, 2nd Ed, Raschka & Mirjalili, PACKT Publishing
The first edition may work, but beware that it contains many typos and errors.
Grading: (Attendance, Team-projects, Midterm, Final) = (10%, 50%, 20%, 20%)
Links
PBL Center: http://pbl.hanyang.ac.kr
Q/As:
Online: trello
Office Hour of TAs: Wed 2~3pm. Cluster Bd. Room 620 (AI LAB)
TA : nomar0107@gmail.com, RA : max2747@naver.com
Exams
Midterm
Date/time: Oct 12 (Friday), 6 ~ 8pm
Location: Conference Hall, room 104 (campus map)
Important Topics [link]
Final
PBL-3 Report is due:
Class 1: Dec 18 by 13:00
Class 2: Dec 20 by 16:00
Email to TA: nomar0107@gmail.com
Subject format (이메일 제목 양식):
Class 1(Tues): [PBL3-1] your student ID number
Class 2(Thurs): [PBL3-2] your student ID number
Send your Jupyter Notebook (pbl3-yourstudentid.ipynb) file as an attachment
YOU MUST FOLLOW THE ABOVE SUBMISSION FORMAT (위 제출 양식을 반드시 지켜야 합니다).
NO LATE SUBMISSION WILL BE ACCEPTED (제출 시한을 넘겨서 도착한 이메일은 0점 처리합니다)
Schedule
Week 1: Introduction to AI [note]
Week 2: Basic methods in ML (Logistic regression, Neural Networks, SVM) [note1, note2]
Week 3~5 (PBL Case Study #1) (How to read MNIST files)
Team composition(class1 , class2), Problem understanding, Per team discussion & solution making
Team presentations & feedbacks (send presentation files to nomar0107@gmail.com)
Cross validation & Grid Search [note]
Team presentations with improvements, Team / Member evaluations
Week 6: Midterm (Oct 12)
Week 7: Gradient-based Learning
Week 8~10 (PBL Case Study #2)
Team composition(class1, class2), Problem understanding, Per team discussion & solution making
Team presentations & feedbacks (class 1: Nov. 6, class 2: Nov. 8)
Your team needs to a short presentation (send slides to TA)
Content:
Understanding of SVM using SGD
Your strategy for coding:
Show your class body (sketch)
If you're going to use GridSearchCV
which hyperparams to tune
feature transformation?
who's going to do what...
SGD Algorithm for solving SVM [note]
Code submission 1 (class 1: by Nov 12, class 2: by Nov 14)
TA : nomar0107@gmail.com
Python code: team#.py <training_data> <test data>
Specify what is your training data.
Fix hyperparameters in your code.
Output labels per line, as explained in PBL 2 problem description.
Unsupervised Learning (PCA)
Additional MNIST data
Code submission 2 (class 1: by Nov 19, class 2: by Nov 21)
TA : nomar0107@gmail.com
Python code: team#.py <training_data> <test data>
Final resentation (class 1: Nov 13, class 2: Nov 15)
Collect all contents from your previous talks: you may only briefly explain them during the final talk.
Details about feature extraction
Details of hyperparameters chosen, with plots representing the difference of accuracy with respect to the hyperparameters
Contribution of team members: who did what?
Team presentations with improvements, Team / Member evaluations
Week 12~14 (PBL Case Study #3)
Week 12: Unsupervised Learning: LDA, Kernel PCA [note]
Weak 13: Clustering, Evalution Metric, Variable selection
Weak 14: Decison tree, Ensemble Methods
Week 15: Activity report, Final Exam (Report)