update papers and news

_bibliography/papers.bib

@@ -1,5 +1,36 @@
 ---
 ---
+@inproceedings{mujkanovic22defenses,
+  author          = {Mujkanovic, Felix and Geisler, Simon and Bojchevski, Aleksandar and G{\"u}nnemann, Stephan},
+  title           = {Are Defenses for Graph Neural Networks Robust?},
+  booktitle       = {Neural Information Processing Systems, {NeurIPS}},
+  year            = {2022},
+  abbr            = {NeurIPS},
+  category        = {conference},
+  abstract        = {A cursory reading of the literature suggests that we made a lot of progress in designing effective adversarial defenses for Graph Neural Networks (GNNs). Yet, the standard methodology has a serious flaw -- virtually all of the defenses are evaluated against non-adaptive attacks leading to overly optimistic robustness estimates. We perform a thorough robustness analysis of 7 of the most popular defenses spanning the entire spectrum of strategies, i.e. aimed at improving the graph, the architecture, or the training. The results are sobering -- most defenses show no or only marginal improvement compared to an undefended baseline. We advocate using custom adaptive attacks as a gold standard and we outline the lessons we learned from successfully designing such attacks. Moreover, our diverse collection of perturbed graphs forms a (black-box) unit test offering a first glance at a model's robustness.},
+  selected        = {true},
+}
+@inproceedings{scholten22defenses,
+  author          = {Scholten, Yan and Schuchardt, Jan and Geisler, Simon and Bojchevski, Aleksandar and G{\"u}nnemann, Stephan},
+  title           = {Randomized Message-Interception Smoothing: Gray-box Certificates for Graph Neural Networks},
+  booktitle       = {Neural Information Processing Systems, {NeurIPS}},
+  year            = {2022},
+  abbr            = {NeurIPS},
+  category        = {conference},
+  abstract        = {Randomized smoothing is one of the most promising frameworks for certifying the adversarial robustness of machine learning models, including Graph Neural Networks (GNNs). Yet, existing randomized smoothing certificates for GNNs are overly pessimistic since they treat the model as a black box, ignoring the underlying architecture. To remedy this, we propose novel gray-box certificates that exploit the message-passing principle of GNNs: We randomly intercept messages and carefully analyze the probability that messages from adversarially controlled nodes reach their target nodes. Compared to existing certificates, we certify robustness to much stronger adversaries that control entire nodes in the graph and can arbitrarily manipulate node features. Our certificates provide stronger guarantees for attacks at larger distances, as messages from farther-away nodes are more likely to get intercepted. We demonstrate the effectiveness of our method on various models and datasets. Since our gray-box certificates consider the underlying graph structure, we can significantly improve certifiable robustness by applying graph sparsification.},
+}
+@inproceedings{geisler22generalization,
+  author          = {Geisler, Simon and Sommer, Johanna and Schuchardt, Jan and Bojchevski, Aleksandar and G{\"u}nnemann, Stephan},
+  title           = {Generalization of Neural Combinatorial Solvers Through the Lens of Adversarial Robustness},
+  booktitle       = {International Conference on Learning Representation, {ICLR}},
+  year            = {2022},
+  abbr            = {ICLR},
+  category        = {conference},
+  abstract        = {End-to-end (geometric) deep learning has seen first successes in approximating the solution of combinatorial optimization problems. However, generating data in the realm of NP-hard/-complete tasks brings practical and theoretical challenges, resulting in evaluation protocols that are too optimistic. Specifically, most datasets only capture a simpler subproblem and likely suffer from spurious features. We investigate these effects by studying adversarial robustness -- a local generalization property -- to reveal hard, model-specific instances and spurious features. For this purpose, we derive perturbation models for SAT and TSP. Unlike in other applications, where perturbation models are designed around subjective notions of imperceptibility, our perturbation models are efficient and sound, allowing us to determine the true label of perturbed samples without a solver. Surprisingly, with such perturbations, a sufficiently expressive neural solver does not suffer from the limitations of the accuracy-robustness trade-off common in supervised learning. Although such robust solvers exist, we show empirically that the assessed neural solvers do not generalize well w.r.t. small perturbations of the problem instance.},
+  html            = {https://openreview.net/forum?id=vJZ7dPIjip3},
+  pdf             = {https://openreview.net/pdf?id=vJZ7dPIjip3},
+  talk            = {https://www.youtube.com/watch?v=HVRpjMtZHCk},
+}
 @inproceedings{geisler2021robustness,
 author = {Geisler, Simon and Schmidt, Thomas, and {\c{S}}irin, Hakan and Z{\"u}gner, Daniel, and Bojchevski, Aleksandar and G{\"u}nnemann, Stephan},
 title = {Robustness of Graph Neural Networks at Scale},

_news/papers_iclr22.md

@@ -0,0 +1,6 @@
+---
+layout: post
+date: 2022-01-30
+inline: true
+---
+Our [paper](/publications#geisler22generalization) on generalization of combinatorial solvers was accepted at ICLR 2022.

_news/papers_neurips22.md

@@ -0,0 +1,6 @@
+---
+layout: post
+date: 2022-09-15
+inline: true
+---
+Two papers on robustness, one on [adaptive attacks](/publications#mujkanovic22defenses) and one on [robustness certificates](/publications#scholten22defenses), were accepted at NeurIPS 2022.

_news/teaching_ws_22_23.md

@@ -0,0 +1,7 @@
+---
+layout: post
+date: 2022-09-09
+inline: true
+---
+
+This semester I am co-teaching the [Elements of Machine Learning](https://cms.cispa.saarland/eml22/){:target="_blank"} lecture with [Prof. Jilles Vreeken](https://people.mmci.uni-saarland.de/~jilles/){:target="_blank"}.

_pages/publications.md

@@ -3,7 +3,7 @@ layout: page
 permalink: /publications/
 title: Publications
 description: publications in reversed chronological order </br> * denotes equal contribution # by categories </br>
-years: [2021, 2020, 2019, 2018, 2017]
+years: [2022, 2021, 2020, 2019, 2018, 2017]
 nav: true
 sort: 3
 ---

_pages/teaching.md

@@ -7,8 +7,10 @@ sort: 4
 ---
 
 ### Lectures
-[Machine Learning](https://www.in.tum.de/daml/lehre/wintersemester-202021/machine-learning/){:target="_blank"}
+- [Elements of Machine Learning](https://cms.cispa.saarland/eml22/){:target="_blank"} (Winter Semester 22/23). Co-teaching with [Prof. Jilles Vreeken](https://people.mmci.uni-saarland.de/~jilles/){:target="_blank"}.
+- [Machine Learning](https://www.in.tum.de/daml/lehre/wintersemester-202021/machine-learning/){:target="_blank"}
 (Winter Semester 20/21). Co-teaching with Prof. Stephan Günnemann.
 
 ### Seminars
-[Trustworthy Graph Neural Networks](https://cms.cispa.saarland/tgnn_ws21/){:target="_blank"} (Winter Semester 21/22).
+- [Classic Contributions to Machine Learning](https://cms.cispa.saarland/peml/){:target="_blank"} (Summer Semester 22). Co-teaching with Rebekka Burkholz.
+- [Trustworthy Graph Neural Networks](https://cms.cispa.saarland/tgnn_ws21/){:target="_blank"} (Winter Semester 21/22).