Course Project

One of the main goals of CSEP 590A is to prepare you to apply state-of-the-art data mining tools and algorithms to an application. If you are interested in research, CSEP 590A will also leave you well-qualified to do data mining research. The class's final project will offer you an opportunity to do exactly this.

Students can (and are strongly encouraged) to work in teams of up to three people. If you have a project of such large scope and ambition that it cannot be done by a team of only three, you may propose doing a project in a team of four. Please sign up with your group information using the Google form.

Your first task is to pick a project topic. If you are looking for project ideas, please come to either Prof. Althoff or the TAs' office hours, and we'd be happy to brainstorm and suggest some project ideas.

In the meantime, there can be three kinds of course projects:

• Experimental evaluation of algorithms and models on an interesting dataset, ideally to derive new insights from the dataset.
• A theoretical project that considers a model or an algorithm and derives a rigorous result about it.
• A scalable implementation of an algorithm for processing massive amounts of data.

Ideally, projects will be a mix of the three types of projects outlined above. As with the reaction paper, the project should contain at least some amount of mathematical analysis, and some experimentation on real or synthetic data.

Many fantastic class projects come from students picking either an application/dataset that they're interested in, or picking some sub-field of data mining that they want to explore more, and working on that as their project. If you haven't worked on a research project before but would like to, you can also use this as an opportunity to try your hand at it. (Just be sure to ask us for help if you're uncertain how to best get started.)

Alternatively, if you're already working on a research project that data mining might be applicable to, then working out how to apply data mining techniques to it will often make a very good project topic. Similarly, if you currently work in industry and have an application on which data mining might help, that could also make a great project.

A very good CSEP 590A project will comprise a publishable or nearly-publishable piece of work. Each year, some number of students continue working on their projects after completing CSEP 590A, and submit their work to a conference or journal.

Projects will be evaluated based on:

• The technical quality of the work: Does the technical material make sense? Are the things tried reasonable? Are the proposed algorithms or applications clever and interesting? Do the authors convey novel insight about the problem and/or algorithms?
• Significance: Did the authors choose an interesting or a "real" problem to work on, or only a small "toy" problem? Is this work likely to be useful and/or have impact?
• The novelty of the work, and the clarity of the write-up.

Lastly, a few words of advice: Many of the best class projects come from students working on topics that they're excited about. So, pick something that you can get excited and passionate about! Be brave rather than timid, and do feel free to propose ambitious things that you're excited about. Finally, if you're not sure what would or would not make a good project, please also feel strongly encouraged to either email us or come to office hours to talk about project ideas.

There are four deliverables (Click the respective deliverable to know more):

• Project proposal (20% of the project grade)
• Project milestone report (20% of the project grade)
• Final project report (50% of the project grade)
• Presenting your work at a virtual project presentation session (10% of the project grade)

You can work in groups of 3 people on the project.

Project proposal

This assignment consists of two parts. The first part is a ~2-page reaction paper to several published research papers, and the second part is a ~2-page proposal for the project you want to pursue for this class. Two parts should be related: your reactions to published papers should inform the project you will work on.

Proposals will be evaluated subjectively based on the following criteria (same criteria as listed in the above section for final project evaluation):

• The technical quality of the work: Does the technical material make sense? Are the things tried reasonable? Are the proposed algorithms or applications clever and interesting? Do the authors convey novel insight about the problem and/or algorithms?
• Significance: Did the authors choose an interesting or a "real" problem to work on, or only a small "toy" problem? Is this work likely to be useful and/or have impact?
• The novelty of the work, and the clarity of the write-up.

Please use the NeurIPS 2019 template as given here or here. If you are not using LaTeX, you may follow the formatting instructions given in Section 2 of the links.

Part 1: Reaction paper component (~2 pages). The course is based on material from the last few years. This means that most of it in form of research papers, which raise lot of interesting issues that have yet to be explored. The goal of the reaction paper section is that students familiarize themselves more in depth with the material covered in class, do reading beyond what was covered in class.

Students will pick at three (or more) papers that are clearly related to course topics (e.g., mentioned in class or office hours; if in doubt check with course staff about the papers you aim to read). These papers should go beyond material that was covered in detail in a lecture (i.e., don't only discuss required readings or the textbook). Students should carefully read the papers and write a short (approximately 2 pages) reaction paper about the content of the chosen papers. You should be thinking beyond what you read, and not just take other people's work for granted. The reaction part of the paper should address the following questions:

• Summary
• What is main technical content of the papers?
• How do papers relate to the topics presented in the course?
• What is the connection between the papers you are discussing?
• Critique
• What are strengths and weaknesses of the papers, and how might they be addressed?
• Do the proofs in the paper seem technically sound?
• Are the theorems simple to understand, and do they indeed convey the meaning that the authors claim?
• What were the authors missing?
• Was anything particularly unrealistic? Do all the assumptions make sense?
• Brainstorming (that then leads to the project proposal)
• What are promising further research questions in the direction of the papers?
• Can we generalize the results in the paper?
• Could we get something interesting by adding or dropping some assumptions?
• How could they be pursued?
• An idea of a better model for something? A better algorithm? A test of a model or algorithm on a dataset or simulated data?

Reaction papers should not just be summaries of the papers you read. The last two bullets should form the most substantial part of the document. Answering these questions can be a very good way to explore a potential project topic. The reaction paper should be concluded with a section with a description of some promising further research directions and questions, and how could they be pursued.

In prior versions of the course, the reaction paper has been a very good way to explore a potential project topic.

Part 2: Project proposal (~2 pages). The project proposal component should build on the reaction paper component. The purpose of the reaction paper is to survey the related work and identify what are strengths and weaknesses of the papers and how they may be addressed. The proposal should then focus on what are some promising further research directions and questions: How precisely do you plan to pursue them? What methods/data do you plan to use? You should try to provide a concrete proposal for a model or algorithm that potentially extends or improves the topics discussed in the papers you've read.

When writing the proposal you should try to answer the following questions:

• What is the problem you are solving?
• What data will you use (how will you get it)?
• What work do you plan to do the project?
• Which algorithms/techniques/models you plan to use/develop? Be as specific as you can!
• How will you evaluate your method? How will you test it? How will you measure success?
• What do you expect to submit/accomplish by the end of the quarter?

Some other points to note:

• The project should contain some mathematical analysis and some experimentation on real or synthetic data
• The result of the project will typically be a 8 page paper, describing the approach, the results, and the related work.

We strongly encourage you to work in groups of 3 people. It is hard for us to balance the grading based on the group size. This means that projects will be graded about the same regardless of how many people are in the group -- working in groups is strongly encouraged!

Project Milestone

• Think of this as a draft of your final report but without your major results.
• We expect that you have completed 50% of the project. For most projects, this would include implementing and evaluating at least a baseline method/model.
• Provide a complete picture of your project even if certain key parts have not yet been implemented/solved.
• We grade mainly based on the quality, as well as the completion of the following sections:
• Thorough introduction of your problem
• Review of the relevant prior work (you already started on this in your proposal)
• Description of the data collection process
• Description of any initial findings/results (e.g. results from the baseline model) and summary statistics from your dataset
• Description of any mathematical background necessary for your problem
• Formal description of any important algorithms used
• Description of general difficulties with your problem which bear elaboration
• An outline of the parts which have not yet been completed so that it is clear specifically what you plan to do for the final version
• Recommended length 3-5 pages. At most 10 pages (excluding references).

Project Report

The final project report should be a 5-10 page paper, describing the introduction, related work, approach, results and conclusion. We will not accept reports longer than 10 pages (page count includes figures, but excludes references). At the end of the report, you should also highlight the contributions of individual team members to the project (in the format outlined below). The project report should contain at least some amount of mathematical analysis, and some experimentation on real or synthetic data.

Course staff will use the following guidelines when grading your final project write-ups. Keep in mind however, that if there is a good reason why your project doesn't match the rubric below, we will take that into consideration when grading your report. For example, we recognize that purely theoretical or pure data analysis projects may not fit the rubric below perfectly, and that depending on your project you may want swap the ordering of certain sections. But hopefully all projects can be roughly mapped to the criteria below:

• Introduction/Motivation/Problem Definition
  • What is it that you are trying to solve/achieve and why does it matter.
• Related Work
  • How does your project relate to previous work. Please give a short summary on each paper you cite and include how it is relevant.
• Data Collection
  • Summary of your data and any preprocessing you might have done with it.
• Model/Algorithm/Method
  • This is where you give a detailed description of your primary contribution. It is especially important that this part be clear and well written so that we can fully understand what you did.
• Results and findings
  • How do you evaluate your solution to whatever empirical, algorithmic, or theoretical question you have addressed? It is not so important how well your method performs but rather how interesting and clever your experiments and analysis are.
  • We are interested in seeing a clear and conclusive set of experiments which successfully evaluate the problem you set out to solve. Even if you have a theoretical project you should have something here to demonstrate the validity or value of your project (for example, proofs or runtime analysis).
• Discussion of results/insights/error analysis
  • Make sure to interpret the results and talk about what can we conclude and learn from your evaluations. What do these evaluation methods tell you about your solution?
  • What does your model/algorithm/method get wrong and why? Are the errors systematic? Why are you willing to accept these errors or what could be done to fix them?
• Technical Quality
  • Does the technical material make sense? Are the things tried reasonable? Are the proposed algorithms or applications clever and interesting? Do the authors convey novel insight about the problem and/or algorithms? Is the work novel in some way?
• Significance
  • Did the authors choose an interesting or a "real" problem to work on, or only a small "toy" problem? Is this work likely to be useful and/or have impact?
• Style and writing
  • Overall writing, grammar, organization and neatness.

Unlike the project proposal and milestone, we plan to assign individual scores to team members for the final project report. We observed that there is a skewed distribution of work in some of the teams and would like to take that into account when we are grading. Your score for the final report will now be a function of two aspects:

1. The criteria outlined above for your final report
2. Your contribution to the project relative to that of your team members.

In order to do be able to assign such individual scores, we want you to write down a brief summary of the individual contributions of each of the team members in the format outlined below at the end of each report:

Example:
----------------
Team Member 1: Plotting graphs during data analysis, crawling the data, preliminary data analysis
Team Member 2: Problem formulation, writing up the report, coming up with the algorithm
Team Member 3: Coding up the algorithm, running tests, tabulating final results
---------------

If you fail to outline individual contributions at the end of your report, we will assign equal score to all the team members.

Virtual Project Presentation Session

The goal of the project presentation session is to give you a chance to see what your classmates have been working on. Instead of an in-person poster session, we will be holding a virtual project presentation session on Zoom. We will divide the project groups into two parallel sessions, each of which will be moderated by different TAs. In each session, we will go through the individual presentations, which consist of a 7-minute prerecorded video, along with a 5-minute Q&A session.

7-minute video

• Please focus on the work that you did (problem statement, implemented algorithms, evaluation criteria), and what you have learned (insights, difficulty).
• We cannot accept any video that is longer than 7 minutes.
• The presentation videos should be uploaded to Gradescope by TBD.
• On the presentation day, the TA hosts will present these videos one after another.

5-minute Q&A session

• All team members are expected to be on the Zoom call for the Q&A session.
• Extra credit for participation. After every presentation, there will be a short Q&A session for students to ask questions through chat. We will give extra credit to students (identified by UW email account) who ask at least one good question, so be sure to pay attention!

Research Project Policy

Many of the students in this course are actively conducting research in data mining and machine learning, so we welcome and even encourage projects that align with research goals beyond this course. However, it is critical that projects define what they will specifically accomplish in the scope of the course. The course project must stand on its own, not merely be a snapshot of an outside research process.

Example Projects

Here are example projects from the CSE 547 version of the course in Spring 2019, 2020 and 2021. For this course, it can but does not have to be network-related, and we especially focus on projects with datasets of non-trivial size. You should not be able to be trivially solve your project problem quickly on your laptop. Your project should involve at least one highly non-trivial component related to this course.

Project Title	Student(s)	Resources
Transformer-based models for protein structure generation	Abhijit Bhatnagar, David Juergens, Prashant Rangarajan	Report
Fake News Detection on Social Media	Anton Lykov, Sandeep Tiwari, Venkata Sai Muktevi	Report
Clustering Covid-19 Viral Genomes and Spike Protein Sequences	Aishwarya Mandyam, Cheng Ni, Jack Khuu	Report
Collaborative Filtering using a Trust Social Network with Deep Learning for Initialization	Weijia Zhang, Reed Zhang, Zhitao Yu	Report
Evaluating Identity and Arguments Online	Amanda Baughan, Elena Khasanova, Erik Tomasic	Report
Gene Expression Data Mining of ARCHS4 Project Proposal	Aleksander Braksator, Qinai Xu, Haobo Zhang	Report
Champion Recommender System For League of Legends	Young Seok Kim, Pradeep Prabhakar, Aniruddha Dutta	Report
Modeling Wellness using Smartphone and Activity Data	Galen Weld, Orson Xu, Ather Sharif	Report
Overlapping Community Detection via Edge-space Representation	Ruojin he, Fengjie Chen, Yikun Zhang	Report
Air Pollution Mapping and Prediction	Mingyu Wang, Su Ye, Gaurav Mahamun	Report
BITSCOPE: Scaling Bitcoin Address De-anonymization using Multi-resolution Chaining	Tianyi Zhou, Zhitong (Mia) Xie, Zhen Zhang	Report
Transformer Based Video Matting	Jackson Stokes	Report