Asja Fischer
Institution: Ruhr-Universität Bochum / CASA
Research Hub(s):
Hub C: Secure Systems
E-Mail: Asja.Fischer@rub.de
Adress: IB 3/153
Website: https://informatik.rub.de/ml/
Publications:
Characteristics of Monte Carlo Dropout in Wide Neural Networks Layer-wise linear mode connectivity Information Plane Analysis for Dropout Neural Networks Wasserstein dropout Marginal Tail-Adaptive Normalizing Flows Uncertainty-Based Detection of Adversarial Attacks in Semantic Segmentation DistriBlock: Identifying adversarial audio samples by leveraging characteristics of the output distribution Benchmarking the Fairness of Image Upsampling Methods Robustifying automatic speech recognition by extracting slowly varying features Copula-Based Normalizing Flows Thresholded Adaptive Validation: Tuning the Graphical Lasso for Graph Recovery Learning Sparse Codes with Entropy-Based ELBOs Uncertainty-Weighted Loss Functions for Improved Adversarial Attacks on Semantic Segmentation SmoothLRP: Smoothing LRP by Averaging over Stochastic Input Variations Approaches to Uncertainty Quantification in Federated Deep Learning Efficient Calculation of Adversarial Examples for Bayesian Neural Networks Detecting Compositionally Out-of-Distribution Examples in Semantic Parsing Detecting Adversarial Examples for Speech Recognition via Uncertainty Quantification A Representative Study on Human Detection of Artificially Generated Media Across Countries AEROBLADE: Training-Free Detection of Latent Diffusion Images Using Autoencoder Reconstruction Error Towards the Detection of Diffusion Model Deepfakes How Sampling Impacts the Robustness of Stochastic Neural Networks Infinite ensembles for uncertainty prediction (Workshop Contribution) Uncertainty quantification with compound density network (Workshop Contribution) Leveraging Frequency Analysis for Deep Fake Image Recognition
