Trustworthy Graph Neural Networks Aleksandar Bojchevski

News

05.04.2022

Grades & Presentation Recordings

Dear Students,

Your grades have been entered into LSF and you should be able to view them. A link to the recordings of the final presentations is available on the Materials page.

Thanks again for your participation in this seminar, your hard work, and the... Read more

Dear Students,

Your grades have been entered into LSF and you should be able to view them. A link to the recordings of the final presentations is available on the Materials page.

Thanks again for your participation in this seminar, your hard work, and the interesting discussions during the final presentations.

 

09.03.2022

Final Presentations

Dear Students,

As previously announced the final presentations are scheduled for this Thursday (10.03) from 13:00 -- 18:00, and this Friday (11.03) from 13:00 -- 18:00. You can use this link to join the Zoom meeting. Remember that attendance is mandatory.

A... Read more

Dear Students,

As previously announced the final presentations are scheduled for this Thursday (10.03) from 13:00 -- 18:00, and this Friday (11.03) from 13:00 -- 18:00. You can use this link to join the Zoom meeting. Remember that attendance is mandatory.

A few small last-minute tips on your presentations (you are allowed to make changes compared to the slides you uploaded):

  • Make sure you have practiced your presentation at least once and that you are within the allotted time limit (25 – 30 minutes)!
  • Your slides should have slide numbers, so we can easily refer to them during the Q&A part.
  • I would avoid having a final slide that is just a "Thank You" or just a "Questions?" slide. Often, this slide stays visible the longest during the Q&A and it's "wasted". Instead, I would make the final slide the summary slide of what you have been talking about. This might inspire people to ask more questions.

See you all tomorrow!

 

25.02.2022

Deadline Extension

Dear Students,

The deadline for submitting the second version of your paper and your presentation has been extended for one week. The new deadline is 04.03.2022 (next Friday).

21.02.2022

Final Presentation Dates & Submission Deadline

Dear Students,

The only two dates when everyone is available according to the poll, and thus the official dates for the final presentations, are:

  • Thursday (10.03) Afternoon (13:00 -- 18:00)
  • Friday (11.03) Afternoon (13:00 -- 18:00)

On Thursday we... Read more

Dear Students,

The only two dates when everyone is available according to the poll, and thus the official dates for the final presentations, are:

  • Thursday (10.03) Afternoon (13:00 -- 18:00)
  • Friday (11.03) Afternoon (13:00 -- 18:00)

On Thursday we will have the presentations for the following topics in the given order:

  • Evasion attacks
  • Poisoning attacks
  • Privacy attacks
  • Privacy-preserving GNNs

On Friday:

  • Bias
  • Algorithmic fairness
  • Instance-level explanations
  • Model-level explanations

Reminder that attendance is mandatory, and that the deadline for submitting the second version of your paper and your presentation is this Friday (25.02.22).

 

15.02.2022

Presentation Dates

Dear Students,

Almost all of you voted to have the final presentations on two days. To make sure that we find a date that is suitable for everyone please mark all of the dates when you are available. As a reminder, attending the final presentations is mandatory!
... Read more

Dear Students,

Almost all of you voted to have the final presentations on two days. To make sure that we find a date that is suitable for everyone please mark all of the dates when you are available. As a reminder, attending the final presentations is mandatory!

For many dates you will see two options:

  • Morning: 09:00 -- 13:00
  • Afternoon: 13:00 -- 18:00

Please select both of them if both fit your schedule. You have until the end of the week to complete the poll.

02.02.2022

LSF Registration

Dear Students,

At the moment only 4 of you have officially registered the seminar on LSF. If you want the credits (and grade) to officially count you need to be registered. Since the registration deadline (19.11) has passed please let me know as soon as possible... Read more

Dear Students,

At the moment only 4 of you have officially registered the seminar on LSF. If you want the credits (and grade) to officially count you need to be registered. Since the registration deadline (19.11) has passed please let me know as soon as possible if you want me to ask the study coordinator to register you.

31.01.2022

Reviews & Next Deadline & Presentations & Survey

Dear Students,

 

Thank you for sending in your reviews. You can find a link to a folder with all of the reviews (sorted per paper) on the materials page. Please read the reviews for your seminar paper carefully and incorporate the feedback from your fellow... Read more

Dear Students,

 

Thank you for sending in your reviews. You can find a link to a folder with all of the reviews (sorted per paper) on the materials page. Please read the reviews for your seminar paper carefully and incorporate the feedback from your fellow students as you see fit.

The next deadline is 25.02.2022. You must submit by then:

  1. The second (and essentially final) version of your seminar paper. Note, while there will be a camera-ready deadline one week after the presentations, the second version that you submit next will determine your grade (40%). The optional camera-ready deadline is just there to potentially incorporate very minor points that might arise during your presentation.
  2. Your final presentation. Your presentation should be 25 30 minutes long. I highly recommended that you do at least one practice run to make sure that the story flows nicely and that you are on time. For simplicity please convert your presentation file to PDF when you submit it. If you use an online tool which doesn't allow you to export a PDF (e.g. Prezi), then just include a link in your submission. Below you will find a few helpful references for making effective presentations. Most importantly: 1 single message per slide!

If you would like to get my feedback on the first version of your paper and/or your presentation please reach out via email and we will arrange an individual meeting. Note, this meeting was changed from mandatory to optional, but I highly recommend it.

As you know, the final presentation will be scheduled sometime between 25.02.22 and 31.03.22. First, let us all collective decide whether you prefer to have all presentations in one day (around 6 hours + breaks), or you prefer to split them across two days (3 hours + breaks each day). Use this poll to express you preference. I will leave the poll open until the of the week.

Finally, you have the opportunity to evaluate this seminar. You can this link to do so by 03.02.2022. Your feedback is highly appreciated.

As always, don't hesitate to get in touch if you have any questions.

References:

  1. How To Speak by Patrick Winston
  2. How to avoid death By PowerPoint by David JP Phillips
28.01.2022

Reviews Submission

Dear Students,

This is a brief reminder that the deadline to submit the reviews is today. As always, please submit the reviews on the submissions page.

22.01.2022

Updated Papers

Dear Students,

Updated versions of two seminar papers have been uploaded:

  • The paper on "Instance-level explanations" for which the previously submitted PDF was corrupted.
  • The paper on "Bias" for which the author mistakenly submitted a wrong... Read more

Dear Students,

Updated versions of two seminar papers have been uploaded:

  • The paper on "Instance-level explanations" for which the previously submitted PDF was corrupted.
  • The paper on "Bias" for which the author mistakenly submitted a wrong version.
19.01.2022

Peer Review

Dear Students,

The peer review part of the seminar is next. On the materials page you will find the link to the Review Assignment document where you will see a list of the seminar papers that were assigned to you, and a link to download them. The deadline to... Read more

Dear Students,

The peer review part of the seminar is next. On the materials page you will find the link to the Review Assignment document where you will see a list of the seminar papers that were assigned to you, and a link to download them. The deadline to submit your reviews is 28.01.22. Note that the the PDF file for the "Instance-level explanations" paper is potentially corrupted and may need to be fixed. The author should contact me to resolve the issue. We will upload a fixed file as soon as possible. If you notice other issues with the submissions or the assignments don't hesitate to get in touch.

General Advice. Start by carefully reading each paper, using the 3 pass approach and taking notes along the way. Think about possible criticism and concrete ways of improving upon the paper as you read it. Give specific, actionable feedback and balance the positive and negative. Be critical, but kind. Please make your review as informative and substantiated as possible. Superficial, uninformed reviews are worse than no review as they may contribute noise to the review process.
 
In general, your review should concentrate on the seminar paper itself, not the presented methods.
 
Some of the questions you might ask during the review are:
  • Did the author give a good overview of the field?
  • Do you feel like you get the main ideas now?
  • Do you understand the methods?
  • Did they make every sentence count or did they just waste your time with empty babbling?
  • Do you know which methods seem promising and which ones were just a sham?
  • Do you know what the current challenges in the field are, what needs to be overcome for the next big thing? Do you maybe even have an idea of how this could be done?
The reviews you write will be considered in your final grade. The reviews you receive do not influence your grade. However, the paper that receives the most positive, convincing, high-quality reviews will receive the participant-decided best paper award.
 
Review Form. Your review should be in in plain text format and should have 6 sections as detailed below.
 
1. Summary.  Demonstrate that you've read and understood the paper in detail by summarizing what you feel are the most important points. Aim for precision and conciseness. This is not the place to critique the paper or to evaluate it's strengths. The authors should generally agree with a well-written summary.
 
2. Detailed Comments. List the strong and weak points of the paper, but also provide further input to whether (and why) you think the strengths or the weaknesses are dominating. For each point, indicate the importance of the point at hand: is this a major (important, critical) strength/weakness, or a minor one?
 
Make sure that you address the following aspects:
  • Clarity: Is the submission clearly written? Is it well organized? If not, please make constructive suggestions for improving its clarity. Does it adequately inform the reader?
  • Literature: Is the related literature appropriately discussed? If not, be specific about what is missing. Note that oftentimes it is a question of judgement of whether a result should be mentioned due to page limits. The must-mention results should be directly relevant to the topic of the paper.
  • Soundness: A paper ideally makes claims, which should be well supported, either by theoretical arguments, by experimental results, or by appropriate references. Either say, the paper is sound, or list the problems. Any problem listed needs a justification. Do not just say that something is incorrect, include an explanation of why you think it is incorrect.
  • Miscellaneous minor issues: List any typos, grammar, and other issues which you view as minor but should be addressed in the final version of the paper.
All other relevant details (general flaws, corrections, style, etc.) and (constructive) criticism and feedback that could help improve the work or its presentation can also go here.
 
3. Overall Score.
  • 1. Outstanding paper: it deserves the best paper award
  • 2. Very good paper: there are only a few minor flaws
  • 3. Borderline paper: but has merits that outweigh flaws
  • 4. Borderline paper: but the flaws may outweigh the merits
  • 5. Below the standard: there are multiple major flaws
  • 6. Unacceptable: there are many wrong or misleading claims
 
4. Score Justification. Beyond what you've written above as strengths and weakness, what were the major considerations that led you to your overall score for this paper?
 
5. Time Spent Reviewing. How much time did you spend reviewing this paper (in hours)?
 
6. Confidence Score.
  • 5. You are absolutely certain about your assessment. You are very familiar with the related work and checked the math/other details carefully.
  • 4. You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission.
  • 3. You are fairly confident in your assessment. It is possible that you did not understand some parts of the submission. Math/other details were not carefully checked.
  • 2. You are willing to defend your assessment, but it is quite likely that you did not understand central parts of the submission. Math/other details were not carefully checked.
  • 1. Your assessment is an educated guess. Math/other details were not carefully checked.
 
How to Write a Good Review. If you want to learn how to write a good review you can have a look at the reviewer tutorial below and the resources listed therein. Keep in mind that these are aimed at a slightly different audience, namely, reviewers evaluating novel submissions at a conference. Nonetheless, a lot of the advice is still quite useful.
 
Resources on Reviewing:

Good luck with you reviews!

12.01.2022

Paper v1 Submission

Dear Students,

This is a reminder that the deadline to submit the paper outline is this Friday 14.01.22 at 23:55. You can submit the outline as a PDF document on the submissions page. Once I receive all submissions I will send you the instructions about the peer... Read more

Dear Students,

This is a reminder that the deadline to submit the paper outline is this Friday 14.01.22 at 23:55. You can submit the outline as a PDF document on the submissions page. Once I receive all submissions I will send you the instructions about the peer review process.

02.01.2022

First Version of the Seminar Paper

Dear Students,
 
I hope that you were able to get some rest during the holiday break.
 
The next deadline (14.01.2022) for the first version of your seminar paper is near. As a reminder, the main body of your seminar paper (excluding references and the... Read more
Dear Students,
 
I hope that you were able to get some rest during the holiday break.
 
The next deadline (14.01.2022) for the first version of your seminar paper is near. As a reminder, the main body of your seminar paper (excluding references and the appendix) should be at least 4 pages long. You are free to make it longer, up to a maximum of 10 pages, if you think it is absolutely necessary, but aim to be parsimonious. For your paper you can use the NeurIPS 2021 LaTeX template. You can download the style files from the following link: https://neurips.cc/Conferences/2021/PaperInformation/StyleFiles. You can ignore and remove the Checklist section (before the Appendix).
 
Some general advice regarding the writing:
  • Focus on getting the main ideas across. The goal is to summarize existing research by either: (i) categorizing and comparing many papers in a mini survey, or (ii) focusing on a few papers and going more in depth. The paper should be able to teach the main ideas in the field to a novice reader. Since the topics are rather large and the seminar paper is short, you will have to condense the knowledge as much as possible. This means you will have to focus on the most important aspects and optimize your wording to explain these in the most compact way that is still comprehensible.
  • Highlight the advantages and disadvantages of different approaches and your criticism. Are there any trade-offs, e.g. between accuracy and robustness, or between running time, memory, and performance? What are the open challenges? Does the approach generalize well? Did they overly simplify the problem? Is the theory solid or is it just based on unrealistic toy problems? Are the experiments convincing (setup, baselines)? Is it really a new model or just a minor modification of an old one? Remember, you only have very little space! So concentrate on the important criticism, don't lose yourself in the details!
  • Make sure there is a consistent story and line of argument. Write short and direct sentences. Make sure there is one message per paragraph.
  • Keep the exposition related to graphs and GNNs brief. Since all topics deal with GNNs you can assume that the reader is familiar with the basics of how a GNN works. However, it does make sense to formally introduce the notation you are using.
  • Formal (mathematical) definitions of the main problem being tackled and the related background concepts are encouraged. Try to keep them concise but precise. You are welcome to include any theorems if they help you to make some point, or if they are particularly illustrative. However, avoid adding them just for the sake of making the paper more "mathy". There is also no need to rewrite existing proofs (just provide a reference), unless you think that the proof technique is interesting and worth sharing.
  • If possible you can consider summarizing the main algorithm (or several similar algorithms) in pseudo-code, or in a single illustrative figure.
  • If you are comparing experimental results from several different papers make sure that the results are comparable (e.g. same train/validation/test setup) or point out the relevant differences.
  • If you did everything right you've now become a little expert in your niche and you've likely developed your own opinion. Therefore, you are welcome to include a short (optional) section with your research ideas for future work. Think about the major weaknesses of previous works and how you could solve them. This can both be general directions and specific ideas.

Important: The first version is not a supposed to be just a rough draft. The paper should be mostly finished at this point and most changes should be based on the comments from your supervisor and your peers. Of course, if you have additional ideas in the later stage you are welcome to add them.

Feel free to get in touch if you have any questions. I wish you a great start to the new year and good luck with your seminar papers!

Best,
Aleksandar

22.11.2021

Paper Outline Submission

Dear Students,

This is a reminder that the deadline to submit the paper outline is this Friday 26.11.21 at 23:55. You can submit the outline as a PDF document on the submissions page. Recall that the outline should be around 0.5 - 1 pages long in a bullet list format.

06.11.2021

Assigned Topics and Next Deadline

Dear Students,

The topic assignment is complete and 9/10 students got their first or second choice. You can visit the same document where you provided you ranking to see which topic was assigned to you (denoted with a highlighted cell).

The next step is to... Read more

Dear Students,

The topic assignment is complete and 9/10 students got their first or second choice. You can visit the same document where you provided you ranking to see which topic was assigned to you (denoted with a highlighted cell).

The next step is to read the seed papers provided on the Topics page and write a short outline of your paper. The outline should be around 0.5 - 1 pages long in a bullet list format. It should provide a first rough sketch of the main story and the questions/models/algorithms/techniques that you want to focus on.

You have around 3 weeks to complete this task. The deadline is 26.11.21 at 23:55. Afterwards, you can optionally request a meeting to discuss your outline. More details on how to submit the outline will follow.

29.10.2021

Materials, Topic Assignment, Registration

Dear students,

Thank you for participating in the kickoff meeting! The recording and the slides are uploaded on the materials page.

On the same page (which you can also access via Information->Materials) you will find the link to the live document where you... Read more

Dear students,

Thank you for participating in the kickoff meeting! The recording and the slides are uploaded on the materials page.

On the same page (which you can also access via Information->Materials) you will find the link to the live document where you need to rank the topics according to your preferences. The deadline for completing the ranking is in one week on 05.11.21 at 23:55.

You can find the seed papers for each topic on the topics page (which you can also access via Information->Topics). Skimming them can help you make an informed decision.

On the materials page, in the GNN Resources section, you can find links to lectures and other resources that provide an introduction to graph neural networks. I would recommend starting with the MLGS lecture, but this is optional and up to you. In case you find some other useful resources please share them and I can add them to the list.

Reminder: Registration on the CMS or allocation via SIC does not automatically imply registration on LSF. Therefore, don't forget to register the seminar on LSF. In case you decide to drop out of the seminar let me know ASAP so a student on the waiting list can join.

Show all
 

Seminar Description:

Graph Neural Networks (GNNs) have emerged as a fundamental building block in many artificial intelligence systems. Even beyond uses where the graph structure is explicit (e.g. social networks), they show impressive performance for general object-oriented perception, representation, and reasoning. In this seminar we will cover GNNs that are not only accurate or efficient, but also robust, privacy-preserving, fair, uncertainty-aware, and explainable. We will explore how GNNs fail w.r.t. these trustworthiness aspects and how to improve them.

Organization (block format):

Each student will receive a few research papers on a single topic which they should carefully read and analyze. Starting from these initial papers they should explore the surrounding literature and summarize their main ideas and findings in a 4-page seminar paper. Students will also participate in a peer-review process where they have to provide constructive feedback on each other's work (1 page review for 3 other papers). Finally, each student will prepare and deliver a presentation about their topic during a block seminar at the end of the semester.

Grade:

  • Seminar paper (40%)
  • Presentation (30%)
  • Reviews (30%, 10% for each review)

Schedule:

  • 29.10.21 Kick-off meeting at the start of the semester (online via Zoom)
  • 05.11.21 Topic ranking and assignment
  • 26.11.21 Paper outline deadline [Optional individual meeting]
  • 14.01.22 Paper v1 deadline [Optional individual meeting]
  • 28.01.22 Peer review deadline
  • 25.02.22 Paper v2 + Presentation deadline [Optional feedback]
  • [25.02.22 31.03.22] Final presentations during a 1 or 2 day seminar meeting [Probably hybrid]
  • [+1 week] Optional camera-ready deadline

List of topics:

  1. Evasion attacks
  2. Poisoning attacks
  3. Heuristic defenses
  4. Adversarial training
  5. Robustness certificates
  6. Privacy attacks
  7. Privacy-preserving GNNs
  8. Bias
  9. Fairness
  10. Instance-level explanations
  11. Model-level explanations
  12. Uncertainty-aware models

For more details and seed papers see the topics page.

You should attend this seminar to:

  • Explore and learn about state-of-the-art research on Graph Neural Networks
  • Improve your scientific writing
  • Improve your presentation skills
  • Participate in a review process akin to international conferences


Privacy Policy | Legal Notice
If you encounter technical problems, please contact the administrators