About me
I am a Machine Learning PhD student at the TU Berlin and ATB Potsdam, working on Explainable AI under the supervision of Marina HΓΆhne and Klaus-Robert MΓΌller. I am working at Understandable Machine Intelligence lab and I am part of BIFOLD graduate school. My work is mainly focused on global Explainable AI methods β I develop methods so that we can comprehend what abstractions and concepts Deep Neural Networks learn, what is the purpose and function of individual neurons within the networks, as well as how these neurons pass information to each other, forming more complex circuits.
News
- 07.02.24 ποΈ Gave a talk βIntorduction to Explainable AI: how do we explain Deep Neural Networksβ at BIFOLD Graduate School
- 23.01.24 π New preprint β βManipulating Feature Visualizations with Gradient Slingshotsβ
- 16.01.24 ποΈ Our βConcept-Based global Explainabilityβ track was accepted to XAI-2024
- 10.01.24 ποΈ Invited talk at BLISS Berlin about our NeurIPS 2023 paper βLabeling Neural Representations with Inverse Recognitionβ
- 14.12.23 π βLabeling Neural Representations with Inverse Recognitionβ was presented at NeurIPS 2023
- 14.11.23 π βVisalizing the Diversity of Representations Learned by Bayesian Neural Networksβ was accepted at TMLR
- 30.09.23 ποΈ Presented βMark My Words: Dangers of Watermarked Images in ImageNetβ at XI-ML workshop at ECAI 2023 in Krakau, Poland
- 03.10.23 ποΈ Invited talk at MunichNLP (Video)
- 22.09.23 π βLabeling Neural Representations with Inverse Recognitionβ was accepted at NeurIPS 2023
- 11.09.23 π Participated in the Weizenbaum BIFOLD Summer School 2023
- 26.08.23 β΅οΈ Sailed around the island of Majorca
- 26.07.23 ποΈ Presented βFinding Spurious Correlations with Function-Semantic Contrast Analysisβ at xAI 2023 in Lisboa, Portugal
- 05.07.23 ποΈ Invited talk βExplainable AI: from local to globalβ at Max DelbrΓΌck Center
- 03.07.23 π βDORA: Exploring Outlier Representations in Deep Neural Networksβ accepted at TMLR
- 24.05.23 π Published my first blog On Mechanical Consciousness
- 01.05.23 ποΈ Presenting 2 papers: βDORA: Exploring Outlier Representations in Deep Neural Networksβ and βMark My Words: Dangers of Watermarked Images in ImageNetβ at ICLR 2023 at TrustML-(Un)Limited workshop in Kigali, Rwanda.
- 28.03.23 π 2 papers accepted at ICLR2023 at TrustML-(Un)Limited workshop
- 23.06.22 π Participated in the Weizenbaum BIFOLD Summer School 2022
- 02.06.22 ποΈ Panel discussion on Fair and Trustworthy AI at HelmholtzAI2022 conference
- 25.02.22 ποΈ Presented βNoiseGrad: Enhancing Explanations by Introducing Stochasticity to Model Weightsβ at AAAi 2022 conference
- 02.12.21 π βNoiseGrad: Enhancing Explanations by Introducing Stochasticity to Model Weightsβ acepted at AAAi 2022 conference
- 18.10.21 ποΈ Invited lecture on Explainable AI at Saint-Petersburg State University Graduate School of Management (Video in Russian)
- 23.05.21 ποΈ Invited talk on βExplaining hidden representationsβ at ODS DataFest 2021 (Video in Russian)
- 01.01.2021 π£ Start of the PhD
Research
Paper
| Code
| Twitter Thread
Paper
| Code
| Twitter Thread
Paper
| Twitter Thread
Paper
Paper
| Code
| Video
Paper
| Code
| Twitter Thread
Paper
| Code
| Twitter Thread
| Newspaper Article (DE)