Biomedical Informatics Lab

2019 Fall DHC5036 Machine Learning in Medicine

Course Description

• Intermediate class for machine learning
• Short introduction to machine learning areas
• In-depth explanation on deep learning
• Deep learning application to medical data
• Flipped learning class
  • All students MUST watch course videos in Stanford CS230 Deep Learning
  • Offline class will be Q&A and in-depth discussion session

Prerequisite

• Programming experience
• Basic knowledge on Python, NumPy, and other necessary libraries for deep learning
• Experience on Jupyter notebook or other IDEs
• Basic knowledge on Probability, Linear Algebra, and Derivate
• https://opentutorials.org/module/3733 (git introduce)
• https://opentutorials.org/module/3746 (gui client sourcetree)
• https://opentutorials.org/module/3736 (gui client tortoisegit)

* Though you don't have any knowledge but want to join this class, you had better review this class before the beginning of this course. That class was the extended version of "Deep Learning for All (모두를 위한 딥러닝)" with more explanation (on both theory and Python programming) for undergraduate students.

Instructor

• Prof. Soo-Yong Shin (sy.shin (at) skku.edu)

TA

• yh.choi.phd@gmail.com

Date & Room

• Monday 9:00 AM ~ 11:50 AM
• Jung Yakyong Hall, Ilwon Building

Reference

• Stanford CS230 Deep Learning (& related Coursera courses)
  • All lecture slides can be downloaded via Syllabus & related Coursera courses
  • All lecture videos can be accessed via Syllabus (Course 1, 2 , 3 ,4, 5 in Coursera) and Lecture
  • You first watch lecture video on Coursera and then video on CS230 of each week

* You can watch the lecture video by logging-in Cousera (without payment)

• Additional references: TensorFlow KR post

Evaluation

• Attendance: 10%
• Quiz: 20%
  • Quiz on each class if necessary (no question or discussion at the class)
• Assignment: 20%
  • Assignments will be given irregularly based on the topic of each class
  • Students should submit an assignment before next class
• Project: 50%

Assignment

• All assignments should be submitted to GitHub (https://classroom.github.com/a/ANfw9sxj)

Project

• Any project but related to medical domain
• Detailed explanation document is mandatory
• Project proposal: 10 / 50 %
• Final demo: 40 / 50 %
• TA will test all submitted source codes. Basic criteria:
  • No source codes: 0
  • Can run the codes without any explain: 95 (You have to prepare README.md)
  • Cannot run the codes. Based on the explain and types of errors: 0 ~ 90
  • Any question? send email to Instructor or ask question on each class

Course Schedule

(course schedule can be changed.)

1. 9/2: Course Introduction & Machine Learning Basics
2. 9/9: Machine Learning in Medicine & Lecture 5
   • Further Reading: Highly recommend to read 의료인공지능 written by 최윤섭 (at least his slides)
3. 9/16: Coursera Neural Networks and Deep Learning Week 1-2
   • Lecture 2 C1M1 & C1M2 in Syllabus
4. 9/23: Coursera Neural Networks and Deep Learning Week 3-4 & Stanford CS230 Lecture 2 (on Lectures)
   • Lecture 3 C1M3 & C1M4
5. 9/30: Coursera Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization Week 1
   • Lecture 4 C2M1
6. 10/7: Coursera Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization Week 2-3
   • Lecture 4 C2M2
7. 10/14: Coursera Structuring Machine Learning Projects Week 1-2 & Standford CS230 Lecture 3 & Lecture 6
   • Lecture 5 C3M1 & C3M2
8. 10/21: Project Proposal
9. 10/28: Coursera Convolutional Neural Networks Week 1-2
   • Lecture 6 C4M1 & C4M2
10. 11/4: Coursera Convolutional Neural Networks Week 3-4
    • Lecture 7 C4M3 & C4M4
11. 11/11: Coursera Sequence Models Week 1
    • Lecture 8 C5M1
12. 11/18: No Class (AMIA 2019)
13. 11/25: Coursera Sequence Models Week 2-3
    • Lecture 8 C5M2 & C5M3
14. 12/2: Review & final questions on projects
    • (Optional) Stanford CS230 Lecture 4 & Lecture 7
      • Explainability vs Interpretability
15. 12/9: Project Presentation
16. 12/16: Project Presentation