arXiv subjects

Robin Degraeve

Publications and source records attributed to Robin Degraeve.

3 recordsLinked to original sources

arXiv2026

Reinforcement Learning Disrupts Gradient-Based Adversarial Optimization

Gradient-based adversarial attacks remain a dominant threat to deep neural networks (DNNs), as they exploit gradient information to efficiently optimize adversarial perturbations. To address this, we investigate whether reinforcement learning (RL) training can disrupt the gradient structure used by attackers by training image classifiers with policy-gradient objectives and epsilon-greedy exploration. Through systematic experiments across CIFAR-10, CIFAR-100, and ImageNet-100 with multiple architectures, we find that RL-trained classifiers significantly disrupt gradient-based adversarial optimization. To explain this, we conduct a comprehensive mechanism analysis using loss landscape visualization, static and dynamic gradient indicators, and predictive entropy. Our analysis reveals that RL acts as an implicit regularizer, producing models with highly unstable gradient directions and smaller gradient magnitudes. This combination makes each PGD step both unreliable in direction and limited in magnitude, causing gradient-based attacks to fail within practical iteration budgets. We further show that combining RL with adversarial training (RL-adv) provides a dual-layer defense operating at two complementary levels: RL degrades gradient information available to attackers (gradient-level defense), while adversarial training strengthens decision boundaries (boundary-level defense). RL-adv achieves the highest robustness across all major attack types evaluated, including gradient-based (PGD, AutoAttack), transfer-based, and query-based attacks, outperforming SL-adv by a significant margin. These findings identify RL-induced gradient disruption as a complementary robustness mechanism and motivate future research on hybrid SL-RL training schedules that combine SL's efficiency with RL's gradient-regularization properties.

cs.LG↗

arXiv2025

RTNinja: A generalized machine learning framework for analyzing random telegraph noise signals in nanoelectronic devices

Random telegraph noise is a prevalent variability phenomenon in nanoelectronic devices, arising from stochastic carrier exchange at defect sites and critically impacting device reliability and performance. Conventional analysis techniques often rely on restrictive assumptions or manual interventions, limiting their applicability to complex, noisy datasets. Here, we introduce RTNinja, a generalized, fully automated machine learning framework for the unsupervised analysis of random telegraph noise signals. RTNinja deconvolves complex signals to identify the number and characteristics of hidden individual sources without requiring prior knowledge of the system. The framework comprises two modular components: LevelsExtractor, which uses Bayesian inference and model selection to denoise and discretize the signal, and SourcesMapper, which infers source configurations through probabilistic clustering and optimization. To evaluate performance, we developed a Monte Carlo simulator that generates labeled datasets spanning broad signal-to-noise ratios and source complexities; across 7000 such datasets, RTNinja consistently demonstrated high-fidelity signal reconstruction and accurate extraction of source amplitudes and activity patterns. Our results demonstrate that RTNinja offers a robust, scalable, and device-agnostic tool for random telegraph noise characterization, enabling large-scale statistical benchmarking, reliability-centric technology qualification, predictive failure modeling, and device physics exploration in next-generation nanoelectronics.

cs.LG↗

arXiv2025

Investigation of Low Frequency Noise in CryoCMOS devices through Statistical Single Defect Spectroscopy

High 1/f noise in CryoCMOS devices is a critical parameter to keep under control in the design of complex circuits for low temperatures applications. Current models predict the 1/f noise to scale linearly with temperature, and gate oxide defects are expected to freeze out at cryogenic temperatures. Nevertheless, it has been repeatedly observed that 1/f noise deviates from the predicted behaviour and that gate oxide defects are still active around 4.2 K, producing random telegraph noise. In this paper, we probe single gate oxide defects in 2500 nMOS devices down to 5 K in order to investigate the origin of 1/f noise in CryoCMOS devices. From our results, it is clear that the number of defects active at cryogenic temperatures resulting in random telegraph noise is larger than at 300 K. Threshold voltage shifts due to charged defects are shown to be exponentially distributed, with different modalities across temperatures and biases: from monomodal at 300 K to trimodal below 100 K. The third mode is interpreted in the framework of percolation theory. By fitting these distributions, it is shown that more than 80% of the detected defects belongs to the oxide bulk. Afterwards, starting from the raw data in time domain, we reconstruct the low frequency noise spectra, highlighting the contributions of defects belonging to different branches and, therefore, to different oxide layers. This analysis shows that, although interface traps and large defects associated with the third mode are the main sources of 1/f noise at 5 K, bulk oxide defects still contribute significantly to low-frequency noise at cryogenic temperatures. Finally, we show that defect time constants and step heights are uncorrelated, proving that elastic tunnelling model for charge trapping is not accurate.

cond-mat.mes-hall↗