Institution profile

Institute of Science and Technology Austria

Academic institutioneurope · at
Official website
Research library234linked papers
Opportunities0open roles
Selected work

Representative Papers

Machine Unlearning using Forgetting Neural Networks

Oct 29, 2024arXiv.org

To address the privacy requirement of compliantly erasing specific user data from AI models without costly retraining, this paper proposes a novel machine unlearning method based on Forgetting Neural Networks (FNNs). We introduce the first learnable and verifiable forgetting framework grounded in FNNs, incorporating four distinct forgetting layers—gated, masked, perturbed, and reconstructed—to emulate cognitive forgetting mechanisms. Empirical evaluation on MNIST and Fashion-MNIST demonstrates that our approach significantly reduces membership inference attack success rates (average reduction >40%) while maintaining model utility, with post-forgetting accuracy degradation under 2%. This work provides an efficient, interpretable, and retraining-free solution for model-level data unlearning, advancing trustworthy AI systems.

3 citationsRead paper

Train-Free Segmentation in MRI with Cubical Persistent Homology

Jan 02, 2024arXiv.org

To address the scarcity of annotated data in MRI segmentation, this paper proposes a training-free, fully unsupervised topological segmentation framework. Methodologically, it leverages cubical persistent homology to extract topological features—such as connected components and voids—from MRI volumes; employs automated threshold selection and spatial localization of representative cycles; and integrates anatomical geometric priors (e.g., spheres, cylinders, circles) to achieve precise segmentation of target structures—including glioblastoma, myocardium, and fetal cortical plate. Its key innovation lies in the first use of spatial coordinates of representative cycles to directly guide segmentation, thereby ensuring interpretability, topological stability, and geometric adaptability. Evaluated across multiple clinical MRI tasks, the method matches state-of-the-art supervised approaches in performance while requiring no labeled data—significantly enhancing robustness and clinical trustworthiness.

2 citations1 influentialRead paper

Improved Accuracy for Private Continual Cardinality Estimation in Fully Dynamic Streams via Matrix Factorization

Jan 05, 2026arXiv.org

This work addresses the challenge of high error in differentially private cardinality estimation under the continual observation model for fully dynamic streams, where small changes in the stream can cause large fluctuations in sensitivity. The paper presents the first systematic characterization of the ℓₚ-sensitivity vector structure inherent to differentially private streaming, and integrates this insight with an optimized counting matrix factorization mechanism to achieve high-accuracy continual estimation in streams supporting both insertions and deletions. The proposed method significantly reduces privacy error for tasks such as distinct counting, degree distribution, and triangle counting, outperforming existing approaches both theoretically and empirically across a broad range of parameter settings.

1 citationsRead paper

Connecting Neural Models Latent Geometries with Relative Geodesic Representations

Jun 02, 2025

Neural models trained on identical tasks exhibit geometrically heterogeneous latent representations due to stochasticity and architectural differences, rendering cross-model representations incomparable. To address this, we propose relative geodesic representations grounded in pullback metrics—a novel application of differential-geometric pullback metrics to latent space alignment—explicitly modeling intrinsic geometric transformations between latent manifolds of distinct models. Unlike conventional linear alignment methods, our approach operates without supervision and generalizes across diverse architectures and pretraining paradigms. Experiments on autoencoders and vision foundation discriminative models demonstrate substantial improvements in cross-model retrieval accuracy and model stitching performance. Moreover, the method scales effectively to large-scale settings.

1 citationsRead paper

Stochastic Gradient Descent over P2

Sep 11, 2026

研究通过将问题提升到线性希尔伯特空间并构造高斯随机场近似，解决了在概率测度上使用随机梯度下降进行优化的问题。

0 citationsRead paper
Recent publications

Latest Papers

Stochastic Gradient Descent over P2

Sep 11, 2026

研究通过将问题提升到线性希尔伯特空间并构造高斯随机场近似，解决了在概率测度上使用随机梯度下降进行优化的问题。

0 citationsRead paper