(EDIT: no longer accepting new participants.)
Looking for participants in a small academic study.
Time Requirement: 60 minutes (purely online)
Compensation: you can choose between 60$ or a 100$ donation to a charity of your choice.
Prerequisites: the only requirement is a completed university course in computer science or statistics, or something comparable (ask me if you're unsure). You do not need to know anything about image processing.
Task Description: Broadly speaking, the study is about understanding how neural networks process images.
Specifically, we will be looking at the final hidden layer of resnet. This layer consists of 512 filters. When the network processes an image, each filter outputs a 7-by-7 matrix of numbers, where each number corresponds to an image region. In the context of the study, this matrix is converted to binary activations via thresholding, so each filter can be seen as outputting a 7-by-7 binary mask over the image. Here is an example:

Insofar as each filter learns one or more human-understandable concepts and activates on image regions related to these concepts, it is possible to understand their behavior and predict their activations ahead of time. If you participate, your job will be to do this for one of the 512 filters that will be assigned to you. You will be given access to a large set of images along with the corresponding masks output by your filter. You can study them to get a sense of what your filter is doing, and you can also test this by hand-drawing masks on the images and comparing them to the activations of your filter. After training like this for 40 minutes, your understanding will be tested on a different sequence of images (you will again have to hand-draw masks predicting the filter's activations). The results will be included in a paper on network interpretability that I'm working on.
If you are interested or have any questions, please contact me via PM or per mail at rafael.harth@gmail.com. (I will give you a full description of the process, then you can decide whether or not you want to participate.) Participation will be limited to at most 6 people.
60 dollars for you or 100 for charity.
That said, registrations are now closed (I've reached the required number of participants a few hours ago and forgot to update this comment).
(
I saw October didn't have one. First post - please let me know if I do something wrong.
To whoever comes after me: Yoav Ravid comments that the wording could use an update.
)
If it’s worth saying, but not worth its own post, here's a place to put it.
If you are new to LessWrong, here's the place to introduce yourself. Personal stories, anecdotes, or just general comments on how you found us and what you hope to get from the site and community are invited. This is also the place to discuss feature requests and other ideas you have for the site, if you don't want to write a full top-level post.
If you want to explore the community more, I recommend reading the Library, checking recent Curated posts, seeing if there are any meetups in your area, and checking out the Getting Started section of the LessWrong FAQ. If you want to orient to the content on the site, you can also check out the new Concepts section.
The Open Thread tag is here. The Open Thread sequence is here.