CS 335
Home
People
Grading
Attendance Policy
Weekly Reading
Textbooks
Resources
Office Hours
Lectures
Weekly Schedule
Past Lectures
Course Project
Guidelines
Datasets & Ideas
Final Reports

CS 335: Fair, Accountable, and Transparent (FAccT) Deep Learning

Spring 2020 Project Guidelines

Your course project is an excellent opportunity to explore techniques learned from class and apply them to practice. Here is a list of project ideas along with datasets to get you started. You are encouraged to discover FAccT related questions in your own datasets.

Project Expectation

We expect the course project to be computational in nature. Students should ideally implement a deep learning based approach to solve a problem related to FAccT. We encourage students to explore innovative ideas that improve existing baselines.

Project Policy

  • Projects can be done either individually or by two students. For group projects, we ask that you divide work equally.

  • You are expected to implement a FAccT-related algorithm, model, or framework onto one and more datasets of your choice. Novel innovations here are strongly encouraged.

  • No double dipping. You are expected to develop new work and can not reuse projects from another course or another research project. However, we do encourage you to take problems or datasets from your own research or a problem that you are familiar with.

  • Your project will contribute to 75% of your final grade, and will consist of deliverables from the following four stages. You will need to follow the NeurIPS template when writing your final report.

    • Project proposal, 1.5 pages (10%), due Apr 22 before class

    • Milestone presentation (5%), May 13

    • Milestone project report, 5 pages (10%), due May 13 before class

    • Final presentation (30%), May 29, Jun 3

    • Final report: maximum of 8 pages excluding references and appendices (45%), due Jun 12 11:59PM PDT

Project Proposal

Your project proposal should clearly illustrate the following. The project proposal is limited to 1.5 pages under the NeurIPS template.

  • the problem that you are solving

  • dataset of your project

  • evaluation metrics of your project

  • baselines of your project

Milestone Presentation

We ask each group to make a quick presentation to illustrate the basic ideas of their projects and get feedback from the class.

Milestone Project Report

Your proejct Milestone report should consist of the progress that you have made so far with a limit of 5 pages. These could include any of the following:

  • your planned/implemented approaches

  • early insights/experimental results

Final Presentation and Final Report

Students are required to make a final presentation to the class to demonstrate their course projects. This is followed by a final project report that consists of a writeup of the student's progress. Final report should be no longer than 8 pages.

Course Progress Check-ins

Our TA will conduct bi-weekly course project check-in with each of the teams. Students are encouraged to meet with TA in office hours to discuss the directions of the project.