COURSE SYLLABUS AND OUTLINE
CSE-5095-004 and Engineering-SEC001-1133
MACHINE LEARNING IN BIOMEDICAL INFORMATICS
| LECTURE:
CAST 204, Mon/Wed 2:00pm — 3:15pm
|
LAB:
ITEB 138, TBD
|
| INSTRUCTOR:
Jinbo Bi Phone: 486-1458 Email: jinbo@engr.uconn.edu Office hours: Tue. 2:30pm — 3:15pm Office: ITEB 233 |
TEACHING ASSISTANT:
Abdulaziz Miyajan Phone: Email: Abdulaziz.miyajan@uconn.edu Office hours: by appointment Office: ITEB |
PURPOSE AND APPROACH:
The purpose of this course is to introduce to the students the general topics and techniques of data mining and machine learning with specific application focus on biomedical informatics. This course introduces multiple real-world medical problems with real patient data, and how multiple analytic algorithms have been used in an integrated fashion to cope with these problems. It covers some cutting-edge data mining technology which can successfully tackle problems that are complex, highly dimensional, and/or ambiguous in labeling. General topics of data mining, such as clustering, classification, regression, dimension reduction, will be described. However, efforts will also be given to more advanced and recent topics. In particular, imprecisely supervised learning problems will be discussed, including multiple instance learning, metric learning, and learning with multi-labeler annotations etc. Throughout the entire course, practical medical/healthcare problems will be used as examples to demonstrate the adoption and effectiveness of data mining methods.
The course will consist of lectures, labs, paper reviews and projects. Lectures will serve as the vehicle to introduce concepts and knowledge to students. Labs will be used to enforce the material given in lectures and students paper reviews will be used to study the state-of-the-art from researchers in the field. Participation is encouraged during the class.
As part of the course, the students will work on a term project with the goal of applying any of the studied techniques to a problem selected from a list of projects. Students are also encouraged to propose and design their own problems which need to be approved by the instructor for class suitability. Teams of two-three students will be created for each project. Each team is required to present in the classroom and submit a project report, of 15-20 pages, which includes the definition of the problem, techniques used to solve the problem and experimental results obtained. This exercise will help the team gain a hands-on understanding of the material studied in this course and promotes collaborations among team members.
COURSE OBJECTIVES:
- Get to know some general topics in biomedical informatics
- Focus on some high-demanding medical informatics problems with hands-on experience of applying data mining techniques
- Equip students with knowledge about the basic concepts of machine learning and the state-of-the-art literature in data mining/machine learning
TEXTBOOKS:
- Introduction to Data Mining by Pang-Ning Tan, Michael Steinbach, and Vipin Kumar, ISBN-10: 0321321367
- Pattern Classification (2nd Edition) by Richard O. Duda, Peter E. Hart and David G. Stork, ISBN-10: 0471056693
- Pattern Recognition and Machine Learning (Information Science and Statistics) by Christopher M. Bishop, ISBN-10: 0387310738
GRADING:
- In-Class Lab Assignments (3): 30%
- Paper review (1): 10%
- Term Project (1): 50% (including 2 progress reports and 1 final report with software package)
- Participation: 10%
TENTATIVE SCHEDULE:
| Week | Date | Lecture Notes | HW | Announcement |
| 1 | 1/23 | Introduction | ||
| 1/28 | General topics in machine learning, review basics | |||
| 2 | 1/30 | Project planning and introduction of previous students projects | ||
| 2/04 | General classification and regression | |||
| 3 | 2/06 | Linear models for regression | ||
| 2/11 | Support vector machines | |||
| 4 | 2/13 | Lab (1): homework assignment on classification | ||
| 2/18 | Medical problem 1: Clinical Decision Support Systems | |||
| 5 | 2/20 | Invited Lecture: EEG analyses for neurophysiological disorders by a psychologist Prof. Chen | ||
| 2/25 | General topics on clustering | |||
| 6 | 2/27 | Hierarchical clustering – traditional techniques | ||
| 3/04 | Spectral clustering – modern techniques | |||
| 7 | 3/06 | Medical problem 2: Cardiac Ultrasound Image Categorization (deal with medical images) | ||
| 3/11 | Invited Lecture: drug discovery by Scientist from Pfizer Pharmaceuticals | |||
| 8 | 3/13 | Lab (2): introduction of Matlab and homework assignment on clustering of cardiac image data | ||
| 3/25 | Student paper review presentation | |||
| 9 | 3/27 | Student paper review presentation | ||
| 4/01 | Student paper review presentation | |||
| 10 | 4/03 | Student paper review presentation | ||
| 4/08 | General topics on dimension reduction | |||
| 11 | 4/10 | Unsupervised dim reduction: PCA, CCA, ICA | ||
| 4/15 | Supervised dim reduction: LASSO, group LASSO, or 1-norm SVM | |||
| 12 | 4/17 | Medical problem 3: Computerized Diagnostic Coding (deal with natural language text data) | ||
| 4/22 | Lab (3): conduct dimension reduction assignments on diagnostic coding data | |||
| 13 | 4/24 | Invited Lecture | ||
| 4/29 | Presentation of final term projects | |||
| 14 | 5/01 | Presentation of final term projects (cont.) | ||
| 5/06 | Presentation of final term projects (cont.) | |||
| 15 | 5/10 | Final Exam Week: No Classes, Make-up Exam, Term project reports are due on Friday |
COURSE POLICY:
- Computers are allowed in classroom for taking notes or any activity related to the current class meeting.
- Assignments must be submitted electronically via HuskyCT. If the assignment is handed in late, 10 credits will be reduced for each additional day.
- Participation in paper review itself will earn 80% credits for each review assignment. Paper review presentation slides need to be turned in via HuskyCT before the class that the presentation is scheduled. The quality of your paper review presentation will be judged by the instructor (10 credits) and scoring of peer students in the class (10 credits).
- Assignments and paper reviews will be graded by the teaching assistant assigned to this course under guidance and consulting of the instructor.
- Final term projects will be graded by the instructor based on the clarity and creativity of the project report and the comparison of final presentation of all teams.
MAKEUP PLAN:
- If a lab assignment or a paper review presentation is missed, there will be a take-home final exam to make up the credits.
- If two of the lab assignments or paper reviews are missed, there will be an additional assignment and a take-home exam to make up each of the two items.
HUSKYCT:
A HuskyCT site has been set up for the class. You can access it by logging in with your NetID and password. You must use HuskyCT for submitting assignments and check it regularly for class materials, grades, problem clarifications, changes in class schedule, and other class announcements.
PROJECT MATERIALS:
Paper Review:
(Please select one paper from the following machine learning / data mining papers to present)
You can check this webpage to see your selection and the papers that have been chosen.
The following papers are from International Conference on Machine Learning 2011-2012.
1. A Co-training Approach for Multi-view Spectral Clustering
Abhishek Kumar, Hal Daume III, University of Maryland
2. Information-Theoretic Co-clustering
Inderjit S. Dhillon, Subramanyam Mallela and Dharmendra S. Modha
3. Learning with Whom to Share in Multi-task Feature Learning
Zhuoliang Kang, Kristen Grauman, Fei Sha
4. Automatic Feature Decomposition for Single View Co-training
Minmin Chen, Kilian Weinberger, Yixin Chen
5. A Unified Probabilistic Model for Global and Local Unsupervised Feature Selection
Yue Guan, Jennifer Dy, Michael Jordan
6. On Random Weights and Unsupervised Feature Learning
Andrew Saxe, pang Wei Koh, Zhenghao Chen, Maneesh Bhand, Bipin Suresh, Andrew Ng
7. The Constrained Weight Space SVM: Learning with Ranked Features
Kevin Small, Byron Wallace, Carla Brodley, Thomas Trikalinos
8. Support Vector Machines as Probabilistic Models
Vojtech Franc, Alexander Zien, Bernhard Scholkopf
9. A Graph-based Framework for Multi-Task Multi-View Learning
Jingrui He, Rick Lawrence
10. TrueLabel + Confusions: A Spectrum of Probabilistic Models in Analyzing Multiple Ratings
Chao Liu, Yi-Min Wang
11. Convex Multitask Learning with Flexible Task Clusters
Wenliang Zhong, James Kwok
12. Multi-level Lasso for Sparse Multi-task Regression
Aurelie Lozano, Grzegorz Swirszcz
13. Output Space Search for Structured Prediction
Janardhan Rao Doppa, Alan Fern, Prasad Tadepalli
14. A Complete Analysis of the l_1,p Group-Lasso
Julia Vogt, Volker Roth
15. Information-Theoretical Learning of Discriminative Clusters for Unsupervised Domain Adaptation
Yuan Shi, Fei Sha
16. A convex relaxation for weakly supervised classifiers
Armand Joulin, Francis Bach
17. Inferring Latent Structure From Mixed Real and Categorical Relational Data
Esther Salazar, Lawrence Carin
18. Learning Task Grouping and Overlap in Multi-task Learning
Abhishek Kumar, Hal Daume III
The following papers are from ACM Special Interest Group on Knowledge Discovery and Data Mining 2011-2012
19. A Sparsity-Inducing Formulation for Evolutionary Co-Clustering
Shuiwang Ji*, Old Dominion Univ; Wenlu Zhang, Old Dominion University; Jun Liu, Siemens Corporate Research at Princeton
20. On Socio-Spatial Group Query for Location-Based Social Networks
De-Nian Yang, Academia Sinica; Chih-Ya Shen*, National Taiwan University; Wang-Chien Lee, Pennsylvania State University; Ming-Syan Chen, NTU
21. Robust Multi-Task Feature Learning
Pinghua Gong*, Tsinghua University; Jieping Ye, Arizona State University; Changshui Zhang, Tsinghua University
22. Unsupervised Feature Selection for Linked Social Media Data
Jiliang Tang*, Arizona State University; Huan Liu, Arizona State University
23. Learning from Crowds in the Presence of Schools of Thought
Yuandong Tian*, Carnegie Mellon University; Jun Zhu, Tsinghua University
24. Event-based Social Networks: Linking the Online and Offline Social Worlds
Xingjie Liu*, The Pennsalvania State Univ; QI HE, IBM Almaden Research Center; Yuanyuan Tian, IBM Almaden Research; Wang-Chien Lee, Pennsylvania State University; John McPherson, IBM Almaden Research Center; Jiawei Han, University of Illinois at Urbana-Champaign
25. On the Semantic Annotation of Places in Location-based Social Networks
Mao Ye, Dong Shou, ; Wang-Chien Lee, ; Peifeng Yin, ; Krzysztof Janowicz,
26. Two-locus association mapping in subquadratic runtime
Panagiotis Achlioptas, ; Bernhard Scholkopf, Max Planck Institute; Karsten Borgwardt, Max Planck Institutes
27. Differentially Private Data Release for Data Mining
Noman Mohammed*, Concordia University; Rui Chen, Concordia University; Benjamin Fung, Concordia University; Mourad Debbabi, Concordia University; Philip Yu, University of Illinois at Chicago
28. Collaborative Topic Models for Recommending Scientific Articles
Chong Wang*, Princeton University; David Blei, Princeton Univ
29. Multi-View Clustering Using Mixture Models in Subspace Projections
Stephan Gunnemann*, Ines Faerber, Thomas Seidl
30. Subspace Correlation Clustering: Finding Locally Correlated Dimensions in Subspace Projections of the Data
Stephan Gunnemann*, Ines Faerber, Kittipat Virochsiri, Thomas Seidl
TOOLS:
Tools that may help with course projects (to be complete)
- Matlab Optimization Toolbox
- SVM_Light (support vector machines)
- LIBSVM (support vector machines)
- Bayesian Knowledge Discoverer (BKD): computer program able to learn Bayesian Belief Networks from databases
- Bayes net toolbox for Matlab
- TSP Demo
- LeNet (neural networks)
- Neural networks demo
- Neural networks flash demo
- GAUL (genetic algorithm)
- Java genetic algorithm demo
- A complete notebook GA
- A system for distributing statistical software, datasets, and information by electronic mail, FTP and WWW
- Tools for mining large databases C5.0 and See5
- Description of the SLIPPER rule learner, that is a system that learns sets of rules from data based on original RIPPER rule learner
- Information about Data Mining and knowledge discovery in Databases
- Clustering Algorithms
ACADEMIC INTEGRITY:
You are expected to adhere to the highest standards of academic honesty. Unless otherwise specified, collaboration on assignments is not allowed. Use of published materials is allowed, but the sources should be explicitly stated in your solutions. Violations will be reviewed and sanctioned according to the University Policy on Academic Integrity. Collaborations among team members are only allowed for the final term projects that are selected.
Academic integrity is the pursuit of scholarly activity free from fraud and deception and is an educational objective of this institution. Academic dishonesty includes, but is not limited to, cheating, plagiarizing, fabricating of information or citations, facilitating acts of academic dishonesty by others, having unauthorized possession of examinations, submitting work for another person or work previously used without informing the instructor, or tampering with the academic work of other students.
DISABILITY STATEMENT:
If you have a documented disability for which you are or may be requesting an accommodation, you are encouraged to contact the instructor and the Center for Students with Disabilities or the University Program for College Students with Learning Disabilities as soon as possible to better ensure that such accommodations are implemented in a timely fashion.
Jinbo Bi 2012/8-2012/12
Last revised: 8/27/2012