Scholar
Matthew Wicker
Google Scholar ID: _0qEDNIAAAAJ
Imperial College London & The Alan Turing Institute
Citations & Impact
Citations
1,162
H-index
16
i10-index
17
Publications
20
Co-authors
30
list available
Publications
Resume (English only)
Academic Achievements
- "Robust Explanation Constraints for Neural Networks", accepted at ICLR 2023 – first provable certificates for robustness of gradient-based explanations, in collaboration with Accenture
- "Tractable Uncertainty for Structure Learning", accepted at ICML 2022 and awarded Best Paper at TPM 2022 – introduced probabilistic circuits for causal structure learning with uncertainty
- "Individual Fairness Guarantees for Neural Networks", accepted for oral presentation at IJCAI 2022 – first global individual fairness certification via MILP
- "Bayesian Inference with Certifiable Adversarial Robustness", accepted at AISTATS 2021 – synthesized BNNs with strong robustness guarantees by combining Bayesian inference and certifiable robustness
- "Gradient-Free Adversarial Attacks for Bayesian Neural Networks", accepted at AABI 2021 – studied gradient-free attack performance on BNNs
- "Robustness of Bayesian Neural Networks to Gradient-Based Attacks", accepted at NeurIPS 2020 – demonstrated BNN posteriors are robust to gradient-based attacks in the limit
- PhD research produced 8 papers on BNN robustness, covering statistical safety in autonomous driving, certifiable Bayesian inference, saliency-based attacks, and adversarial examples in 3D deep learning
- Open-source repository: https://github.com/matthewwicker/deepbayes
