arXiv subjects

Georgios Sotiropoulos

Publications and source records attributed to Georgios Sotiropoulos.

3 recordsLinked to original sources

arXiv2026

RTL Fault Injection of a Deployed Graph Neural Network Trigger for Belle II

As particle physics detectors grow in scale, High Energy Physics experiments must process ever-increasing data volumes. Level-1 trigger systems, implemented on Field-Programmable Gate Arrays and increasingly using neural-network algorithms, filter this data in real time. However, their proximity to the interaction point exposes them to radiation, which can corrupt outputs, stall processing pipelines, or damage hardware, with significant financial and scientific consequences. In this work, we present the first Register Transfer Level fault-injection study of a deployed Level-1 hardware neural-network trigger, GNN-ETM in the Belle II trigger system. We target three failure modes most consequential to a real-time trigger pipeline: deadlocks, timeouts, and packet-integrity violations. Through two complementary campaigns, we inject 1 442 840 Single-Event Upsets across 211 245 signals. We find a monitoring asymmetry in the existing verification infrastructure and propose inter-stage liveness monitoring as a more accurate alternative to output-only observation, showing that Mean Time To Failure estimates from the two approaches differ by up to 78.7%. The resulting per-stage data identifies the highest-priority hardening targets.

hep-ex↗

arXiv2026

Contrastive Regularization of Machine Learning Potentials

Machine learning interatomic potentials are trained to predict energies and forces but built to be sampled: their purpose is to drive molecular simulations whose observables average over the equilibrium distribution the potential defines. They exemplify a broader AI problem -- learned regressors deployed as generators -- where pointwise accuracy does not guarantee a correct distribution. We show that potentials trained by standard Mean Squared Error (MSE) minimization on Density Functional Theory (DFT) data can reach chemical accuracy on held-out data, yet still fail as samplers: their trajectories drift into spurious low-energy minima and return thermodynamic observables that depart sharply from the reference. To correct this, we introduce Contrastive Regularized MSE (CRMSE), a post-training step that augments the MSE with a contrastive term derived from the Kullback--Leibler divergence between the potential's implicit Boltzmann distribution and the target. The network serves as its own energy-based model: persistent Langevin chains expose the configurations it drifts into and raise their energy, adding no new ab initio data. On the ethanol and aspirin molecules of the MD17 dataset, CRMSE confines the sampler to the physical basin and recovers the energy distribution, interatomic-distance distributions, and dihedral free-energy profiles to near-quantitative agreement with DFT, while preserving force accuracy and keeping energy errors within chemical accuracy; it remains effective when the training set is sharply reduced. That MSE training fails this way on MD17 -- one of the most widely used benchmarks -- while a minimal contrastive correction repairs it suggests that reliable sampling depends less on data volume than on training the model against the distribution it produces: distribution-level training is not a refinement of regression accuracy, but a distinct requirement.

physics.chem-ph↗

arXiv2025

On the Efficacy of Shorting Corporate Bonds as a Tail Risk Hedging Solution

United States (US) IG bonds typically trade at modest spreads over US Treasuries, reflecting the credit risk tied to a corporation's default potential. During market crises, IG spreads often widen and liquidity tends to decrease, likely due to increased credit risk (evidenced by higher IG Credit Default Index spreads) and the necessity for asset holders like mutual funds to liquidate assets, including IG credits, to manage margin calls, bolster cash reserves, or meet redemptions. These credit and liquidity premia occur during market drawdowns and tend to move non-linearly with the market. The research herein refers to this non-linearity (during periods of drawdown) as downside convexity, and shows that this market behavior can effectively be captured through a short position established in IG Exchange Traded Funds (ETFs). The following document details the construction of three signals: Momentum, Liquidity, and Credit, that can be used in combination to signal entries and exits into short IG positions to hedge a typical active bond portfolio (such as PIMIX). A dynamic hedge initiates the short when signals jointly correlate and point to significant future hedged return. The dynamic hedge removes when the short position's predicted hedged return begins to mean revert. This systematic hedge largely avoids IG Credit drawdowns, lowers absolute and downside risk, increases annualised returns and achieves higher Sortino ratios compared to the benchmark funds. The method is best suited to high carry, high active risk funds like PIMIX, though it also generalises to more conservative funds similar to DODIX.

q-fin.PM↗