arXiv subjects

Ngoc Phu Doan

Publications and source records attributed to Ngoc Phu Doan.

4 recordsLinked to original sources

arXiv2026

On the second-order optimization for spiking neural networks

Spiking Neural Networks (SNNs) offer an energy-efficient alternative to conventional neural networks by exploiting sparse, binary spikes, and event-driven computation. However, the training of SNNs remains challenging, as spiking activations create a sharp loss landscape that hinders training, and diagonal-curvature optimizers such as the Adam family may fail to capture this geometry. The extension of curvature-based optimization methods to SNNs is further complicated by the sparse, discrete, and temporally recurrent nature of their underlying dynamics. To address these limitations, we propose SpiKFAX, a second-order optimization method that formulates a computationally tractable, Kronecker-factored approximation of the Fisher information matrix specifically adapted to the structure of SNNs. Empirical evaluation across five architectures and seven datasets demonstrates that SpiKFAX consistently yields improvements in test accuracy and training stability relative to other popular optimizers.

cs.LG↗

arXiv2026

Batch Normalization Amplifies Memorization and Privacy Risks

Batch Normalization (BN) is widely adopted to enable faster convergence and more stable training of deep neural networks. However, its impact on privacy and memorization has remained largely unexplored. In this work, we investigate the effect of BN layers on the memorization of atypical or outlier samples and its implications for privacy leakage. We conduct an extensive empirical study using three complementary approaches: (i) unintended memorization of out-of-distribution samples, (ii) per-sample influence, and (iii) susceptibility to membership inference attacks (MIA). Across multiple datasets and architectures, we consistently observe that BN substantially increases the memorization of outliers compared to models without BN. Critically, this amplified memorization translates directly into privacy vulnerabilities: models with BN exhibit significantly higher susceptibility to MIAs. We complement our empirical findings with a mechanistic analysis under the exact BN backward pass, which shows that BN amplifies the per-step margin growth of outlier samples during training. Our results highlight an underappreciated privacy risk associated with BN and provide both practical and theoretical insights into how normalization layers can amplify the influence of rare or sensitive training examples.

cs.LG↗

arXiv2026

MALLORN: Many Artificial LSST Lightcurves based on Observations of Real Nuclear transients

The Vera C. Rubin Observatory's 10-Year Legacy Survey of Space and Time (LSST) is expected to produce a hundredfold increase in the number of transients we observe. However, there are insufficient spectroscopic resources to follow up on all of the wealth of targets that LSST will provide. As such it is necessary to be able to prioritise objects for followup observations or inclusion in sample studies based purely on their LSST photometry. We are particularly keen to identify tidal disruption events (TDEs) with LSST. TDEs are immensely useful for determining black hole parameters and probing our understanding of accretion physics. To assist in these efforts, we present the Many Artificial LSST Lightcurves based on the Observations of Real Nuclear transients (MALLORN) data set and the corresponding classifier challenge for identifying TDEs. MALLORN comprises 10178 simulated LSST light curves, constructed from real Zwicky Transient Facility (ZTF) observations of 64 TDEs, 727 nuclear supernovae and 1407 AGN with spectroscopic labels using Gaussian process fitting, empirically-motivated spectral energy distributions from SNCosmo and the baseline from the Rubin Survey Simulator. Our novel approach can be easily adapted to simulate transients for any photometric survey using observations from another, requiring only the limiting magnitudes and an estimate of the cadence of observations. The MALLORN Astronomical Classification Challenge, launched on Kaggle on 15/10/2025, will allow competitors to test their photometric classifiers on simulated LSST data to find TDEs and improve upon their capabilities prior to the start of LSST.

astro-ph.HE↗

arXiv2026

Loss Landscape Poisoning: Targeted Extraction of Unseen Training Data from LLMs

Large Language Models are increasingly trained on proprietary or sensitive data, from private healthcare and financial records to user conversations containing secrets. Ensuring the privacy of such data against extraction attacks has become a central concern. In this paper, we ask whether an attacker who can poison a portion of the training data can facilitate the leakage of a separate target record they have no access to. We answer in the affirmative and show that such leakage can be induced by a poisoning mechanism that reshapes the model's local loss landscape around the target completion. Our key insight is that poisoning to create a sharp loss minimum at the target, surrounded by elevated loss on nearby alternatives, forces the model to memorize the target as the unique low-loss solution in its neighborhood. The attack requires no architectural changes, and generalizes across centralized and federated learning settings. We demonstrate that the attack amplifies privacy leakage across language (up to 100% successful extraction), and vision-language models (up 90% successful extraction). We show that the attack is thwarted when the model is trained to be differentially private. However, we introduce a new attack that directly probes the loss landscape bypassing even differential privacy defenses.

cs.CR↗