Small Projects

Choose any three (due by 10/16)

Term Project

You will be working on a research project that leverages program analysis to address deep learning problems or vice versa (using deep learning to address program analysis problems such as debugging, testing or security). The former is preferred.

By 10/7, you are supposed to come up with the project proposal. You are welcome to discuss with the instructor or the TA before that. You will give ten minutes proposal presentation in the week of 10/7. The presentation will (1) define the problem you are working on; (2) the state-of-the-art; (3) design of your solution; (4) a brief timeline; (5) experiments you plan to do, including the dataset and models you will consider, and the results you plan to collect.

In the last week (12/2), you will give a 15-20 mins project presentation and demo. In the presentation, you will (1) define the problem you work on; (2) explain the state-of-the-art; (3) design details, including the technical challenges you addressed; (4) report the experimental results; (5) give a demo.

From 10/21, in each class meeting, multiple 5-minute slots will be given to projects to report progress and discuss problems.

  • Potential project ideas
    • Feature-space back-door attacks
    • Detecting back-doors
    • Functional testing AI models
    • Black-box attacks
    • Autonomizing programs
    • Proving some property of an AI model
    • Adversarial attack samples generation using symbolic execution
    • Explaining adversarial sample attacks
    • AI bugs classification
projects.txt · Last modified: 2019/09/10 22:05 by 590ta
Recent changes RSS feed Creative Commons License Donate Driven by DokuWiki