arXiv subjects

Kieran Parsons

Publications and source records attributed to Kieran Parsons.

At least 19 recordsLinked to original sources

arXiv2026

EDRP: Enhanced Dynamic Relay Point Protocol for Data Dissemination in Multi-hop Wireless IoT Networks

Emerging IoT applications are transitioning from battery-powered to grid-powered nodes. DRP, a contention-based data dissemination protocol, was developed for these applications. Traditional contention-based protocols resolve collisions through control packet exchanges, significantly reducing goodput. DRP mitigates this issue by employing a distributed delay timer mechanism that assigns transmission-start delays based on the average link quality between a sender and its children, prioritizing highly connected nodes for early transmission. However, our in-field experiments reveal that DRP is unable to accommodate real-world link quality fluctuations, leading to overlapping transmissions from multiple senders. This overlap triggers CSMA's random back-off delays, ultimately degrading the goodput performance. To address these shortcomings, we first conduct a theoretical analysis that characterizes the design requirements induced by real-world link quality fluctuations and DRP's passive acknowledgments. Guided by this analysis, we design EDRP, which integrates two novel components: (i) Link-Quality Aware CSMA (LQ-CSMA) and (ii) a Machine Learning-based Block Size Selection (ML-BSS) algorithm for rateless codes. LQ-CSMA dynamically restricts the back-off delay range based on real-time link quality estimates, ensuring that nodes with stronger connectivity experience shorter delays. ML-BSS algorithm predicts future link quality conditions and optimally adjusts the block size for rateless coding, reducing overhead and enhancing goodput. In-field evaluations of EDRP demonstrate an average goodput improvement of 39.43\% than the competing protocols.

cs.NI↗

arXiv2025

Winning Big with Small Models: Knowledge Distillation vs. Self-Training for Reducing Hallucination in Product QA Agents

The deployment of Large Language Models (LLMs) in customer support is constrained by hallucination (generating false information) and the high cost of proprietary models. To address these challenges, we propose a retrieval-augmented question-answering (QA) pipeline and explore how to balance human input and automation. Using a dataset of questions about a Samsung Smart TV user manual, we demonstrate that synthetic data generated by LLMs outperforms crowdsourced data in reducing hallucination in finetuned models. We also compare self-training (fine-tuning models on their own outputs) and knowledge distillation (fine-tuning on stronger models' outputs, e.g., GPT-4o), and find that self-training achieves comparable hallucination reduction. We conjecture that this surprising finding can be attributed to increased exposure bias issues in the knowledge distillation case and support this conjecture with post hoc analysis. We also improve robustness to unanswerable questions and retrieval failures with contextualized "I don't know" responses. These findings show that scalable, cost-efficient QA systems can be built using synthetic data and self-training with open-source models, reducing reliance on proprietary tools or costly human annotations.

cs.CL↗

arXiv2025

Smoothed Embeddings for Robust Language Models

Improving the safety and reliability of large language models (LLMs) is a crucial aspect of realizing trustworthy AI systems. Although alignment methods aim to suppress harmful content generation, LLMs are often still vulnerable to jailbreaking attacks that employ adversarial inputs that subvert alignment and induce harmful outputs. We propose the Randomized Embedding Smoothing and Token Aggregation (RESTA) defense, which adds random noise to the embedding vectors and performs aggregation during the generation of each output token, with the aim of better preserving semantic information. Our experiments demonstrate that our approach achieves superior robustness versus utility tradeoffs compared to the baseline defenses.

cs.LG↗

arXiv2024

Exploring User-level Gradient Inversion with a Diffusion Prior

We explore user-level gradient inversion as a new attack surface in distributed learning. We first investigate existing attacks on their ability to make inferences about private information beyond training data reconstruction. Motivated by the low reconstruction quality of existing methods, we propose a novel gradient inversion attack that applies a denoising diffusion model as a strong image prior in order to enhance recovery in the large batch setting. Unlike traditional attacks, which aim to reconstruct individual samples and suffer at large batch and image sizes, our approach instead aims to recover a representative image that captures the sensitive shared semantic information corresponding to the underlying user. Our experiments with face images demonstrate the ability of our methods to recover realistic facial images along with private user attributes.

cs.LG↗

arXiv2024

Analyzing Inference Privacy Risks Through Gradients in Machine Learning

In distributed learning settings, models are iteratively updated with shared gradients computed from potentially sensitive user data. While previous work has studied various privacy risks of sharing gradients, our paper aims to provide a systematic approach to analyze private information leakage from gradients. We present a unified game-based framework that encompasses a broad range of attacks including attribute, property, distributional, and user disclosures. We investigate how different uncertainties of the adversary affect their inferential power via extensive experiments on five datasets across various data modalities. Our results demonstrate the inefficacy of solely relying on data aggregation to achieve privacy against inference attacks in distributed learning. We further evaluate five types of defenses, namely, gradient pruning, signed gradient descent, adversarial perturbations, variational information bottleneck, and differential privacy, under both static and adaptive adversary settings. We provide an information-theoretic view for analyzing the effectiveness of these defenses against inference from gradients. Finally, we introduce a method for auditing attribute inference privacy, improving the empirical estimation of worst-case privacy through crafting adversarial canary records.

cs.LG↗

arXiv2024

Variational Randomized Smoothing for Sample-Wise Adversarial Robustness

Randomized smoothing is a defensive technique to achieve enhanced robustness against adversarial examples which are small input perturbations that degrade the performance of neural network models. Conventional randomized smoothing adds random noise with a fixed noise level for every input sample to smooth out adversarial perturbations. This paper proposes a new variational framework that uses a per-sample noise level suitable for each input by introducing a noise level selector. Our experimental results demonstrate enhancement of empirical robustness against adversarial attacks. We also provide and analyze the certified robustness for our sample-wise smoothing method.

cs.LG↗

arXiv2024

Efficient Differentially Private Fine-Tuning of Diffusion Models

The recent developments of Diffusion Models (DMs) enable generation of astonishingly high-quality synthetic samples. Recent work showed that the synthetic samples generated by the diffusion model, which is pre-trained on public data and fully fine-tuned with differential privacy on private data, can train a downstream classifier, while achieving a good privacy-utility tradeoff. However, fully fine-tuning such large diffusion models with DP-SGD can be very resource-demanding in terms of memory usage and computation. In this work, we investigate Parameter-Efficient Fine-Tuning (PEFT) of diffusion models using Low-Dimensional Adaptation (LoDA) with Differential Privacy. We evaluate the proposed method with the MNIST and CIFAR-10 datasets and demonstrate that such efficient fine-tuning can also generate useful synthetic samples for training downstream classifiers, with guaranteed privacy protection of fine-tuning data. Our source code will be made available on GitHub.

cs.LG↗

arXiv2024

AutoHLS: Learning to Accelerate Design Space Exploration for HLS Designs

High-level synthesis (HLS) is a design flow that leverages modern language features and flexibility, such as complex data structures, inheritance, templates, etc., to prototype hardware designs rapidly. However, exploring various design space parameters can take much time and effort for hardware engineers to meet specific design specifications. This paper proposes a novel framework called AutoHLS, which integrates a deep neural network (DNN) with Bayesian optimization (BO) to accelerate HLS hardware design optimization. Our tool focuses on HLS pragma exploration and operation transformation. It utilizes integrated DNNs to predict synthesizability within a given FPGA resource budget. We also investigate the potential of emerging quantum neural networks (QNNs) instead of classical DNNs for the AutoHLS pipeline. Our experimental results demonstrate up to a 70-fold speedup in exploration time.

cs.AR↗

arXiv2024

Why Does Differential Privacy with Large Epsilon Defend Against Practical Membership Inference Attacks?

For small privacy parameter $ε$, $ε$-differential privacy (DP) provides a strong worst-case guarantee that no membership inference attack (MIA) can succeed at determining whether a person's data was used to train a machine learning model. The guarantee of DP is worst-case because: a) it holds even if the attacker already knows the records of all but one person in the data set; and b) it holds uniformly over all data sets. In practical applications, such a worst-case guarantee may be overkill: practical attackers may lack exact knowledge of (nearly all of) the private data, and our data set might be easier to defend, in some sense, than the worst-case data set. Such considerations have motivated the industrial deployment of DP models with large privacy parameter (e.g. $ε\geq 7$), and it has been observed empirically that DP with large $ε$ can successfully defend against state-of-the-art MIAs. Existing DP theory cannot explain these empirical findings: e.g., the theoretical privacy guarantees of $ε\geq 7$ are essentially vacuous. In this paper, we aim to close this gap between theory and practice and understand why a large DP parameter can prevent practical MIAs. To tackle this problem, we propose a new privacy notion called practical membership privacy (PMP). PMP models a practical attacker's uncertainty about the contents of the private data. The PMP parameter has a natural interpretation in terms of the success rate of a practical MIA on a given data set. We quantitatively analyze the PMP parameter of two fundamental DP mechanisms: the exponential mechanism and Gaussian mechanism. Our analysis reveals that a large DP parameter often translates into a much smaller PMP parameter, which guarantees strong privacy against practical MIAs. Using our findings, we offer principled guidance for practitioners in choosing the DP parameter.

cs.CR↗

arXiv2023

Stabilizing Subject Transfer in EEG Classification with Divergence Estimation

Classification models for electroencephalogram (EEG) data show a large decrease in performance when evaluated on unseen test sub jects. We reduce this performance decrease using new regularization techniques during model training. We propose several graphical models to describe an EEG classification task. From each model, we identify statistical relationships that should hold true in an idealized training scenario (with infinite data and a globally-optimal model) but that may not hold in practice. We design regularization penalties to enforce these relationships in two stages. First, we identify suitable proxy quantities (divergences such as Mutual Information and Wasserstein-1) that can be used to measure statistical independence and dependence relationships. Second, we provide algorithms to efficiently estimate these quantities during training using secondary neural network models. We conduct extensive computational experiments using a large benchmark EEG dataset, comparing our proposed techniques with a baseline method that uses an adversarial classifier. We find our proposed methods significantly increase balanced accuracy on test subjects and decrease overfitting. The proposed methods exhibit a larger benefit over a greater range of hyperparameters than the baseline method, with only a small computational cost at training time. These benefits are largest when used for a fixed training period, though there is still a significant benefit for a subset of hyperparameters when our techniques are used in conjunction with early stopping regularization.

cs.LG↗

arXiv2022

Variational Quantum Compressed Sensing for Joint User and Channel State Acquisition in Grant-Free Device Access Systems

This paper introduces a new quantum computing framework integrated with a two-step compressed sensing technique, applied to a joint channel estimation and user identification problem. We propose a variational quantum circuit (VQC) design as a new denoising solution. For a practical grant-free communications system having correlated device activities, variational quantum parameters for Pauli rotation gates in the proposed VQC system are optimized to facilitate to the non-linear estimation. Numerical results show that the VQC method can outperform modern compressed sensing techniques using an element-wise denoiser.

eess.SP↗

arXiv2022

Learning to Learn Quantum Turbo Detection

This paper investigates a turbo receiver employing a variational quantum circuit (VQC). The VQC is configured with an ansatz of the quantum approximate optimization algorithm (QAOA). We propose a 'learning to learn' (L2L) framework to optimize the turbo VQC decoder such that high fidelity soft-decision output is generated. Besides demonstrating the proposed algorithm's computational complexity, we show that the L2L VQC turbo decoder can achieve an excellent performance close to the optimal maximum-likelihood performance in a multiple-input multiple-output system.

eess.SP↗

arXiv2022

Mobility, Communication and Computation Aware Federated Learning for Internet of Vehicles

While privacy concerns entice connected and automated vehicles to incorporate on-board federated learning (FL) solutions, an integrated vehicle-to-everything communication with heterogeneous computation power aware learning platform is urgently necessary to make it a reality. Motivated by this, we propose a novel mobility, communication and computation aware online FL platform that uses on-road vehicles as learning agents. Thanks to the advanced features of modern vehicles, the on-board sensors can collect data as vehicles travel along their trajectories, while the on-board processors can train machine learning models using the collected data. To take the high mobility of vehicles into account, we consider the delay as a learning parameter and restrict it to be less than a tolerable threshold. To satisfy this threshold, the central server accepts partially trained models, the distributed roadside units (a) perform downlink multicast beamforming to minimize global model distribution delay and (b) allocate optimal uplink radio resources to minimize local model offloading delay, and the vehicle agents conduct heterogeneous local model training. Using real-world vehicle trace datasets, we validate our FL solutions. Simulation shows that the proposed integrated FL platform is robust and outperforms baseline models. With reasonable local training episodes, it can effectively satisfy all constraints and deliver near ground truth multi-horizon velocity and vehicle-specific power predictions.

cs.LG↗

arXiv2022

Line-field Coherent Sensing with LED Illumination

We describe a method of low-coherence interferometry based optical profilometry using standard light-emitting diode (LED) illumination and complementary metal-oxide-semiconductor (CMOS) image sensors. A line-field illumination strategy allows for the simultaneous measurement of many points in space. Micron scale accuracy and resolution are achieved and demonstrated using a variety of targets.

physics.optics↗

arXiv2021

DNN-assisted optical geometric constellation shaped PSK modulation for PAM4-to-QPSK format conversion gateway node

An optical gateway to convert four-level pulse amplitude modulation to quadrature phase shift keying modulation format having shaping gain was proposed for flexible intensity to phase mapping which exploits non-uniform phase noise. The power consumption of the optical modulation format conversion can save by making a DNN-based decision on the receiver side for the generated QPSK signal with non-uniform phase noise. A proof-of-principle experiment has shown that an optically geometric constellation shaped QPSK modulated signals generated from regular PAM4 signals with Gaussian-distributed noise. The shaped QPSK signal shows BER and generalized mutual information improvement by 1dB gain through the use of digital neural network signal recovery.

eess.SP↗

arXiv2020

Huffman-Coded Sphere Shaping for Extended-Reach Single-Span Links

Huffman-coded sphere shaping (HCSS) is an algorithm for finite-length probabilistic constellation shaping, which provides nearly optimal energy efficiency at low implementation complexity. In this paper, we experimentally study the nonlinear performance of HCSS employing dual-polarization 64-ary quadrature amplitude modulation (DP-64QAM) in an extended reach single-span link comprising 200 km of standard single mode fiber (SSMF). We investigate the effects of shaping sequence length, dimensionality of symbol mapping, and shaping rate. We determine that the naïve approach of Maxwell-Boltzmann distribution matching - which is optimal in the additive white Gaussian noise channel - provides a maximum achievable information rate gain of 0.18 bits/4D-symbol in the infinite length regime. Conversely, HCSS can achieve a gain of 0.37 bits/4Dsymbol using amplitude sequence lengths of 32, which may be implemented without multiplications, using integer comparison and addition operations only. Coded system performance, with a net data rate of approximately 425 Gb/s for both shaped and uniform inputs, is also analyzed.

eess.SP↗

arXiv2020

Analysis of Nonlinear Fiber Interactions for Finite-Length Constant-Composition Sequences

In order to realize probabilistically shaped signaling within the probabilistic amplitude shaping (PAS) framework, a shaping device outputs sequences that follow a certain nonuniform distribution. In case of constant-composition (CC) distribution matching (CCDM), the sequences differ only in the ordering of their constituent symbols, whereas the number of occurrences of each symbol is constant in every output block. Recent results by Amari \textit{et al.} have shown that the CCDM block length can have a considerable impact on the effective signal-to-noise ratio (SNR) after fiber transmission. So far, no explanation for this behavior has been presented. Furthermore, the block-length dependence of the SNR seems not to be fully aligned with previous results in the literature. This paper is devoted to a detailed analysis of the nonlinear fiber interactions for CC sequences. We confirm in fiber simulations the inverse proportionality of SNR with CCDM block length and present two explanations. The first one, which only holds in the short-length regime, is based on how two-dimensional symbols are generated from shaped amplitudes in the PAS framework. The second, more general explanation relates to an induced shuffling within a sequence, or equivalently a limited concentration of identical symbols, that is an inherent property for short CC blocks, yet not necessarily present for long blocks. This temporal property results in weaker nonlinear interactions, and thus higher SNR, for short CC sequences. For a typical multi-span fiber setup, the SNR difference is numerically demonstrated to be up to 0.7dB. Finally, we evaluate a heuristic figure of merit that captures the number of runs of identical symbols in a concatenation of several CC sequences. For moderate block lengths up to approximately 100 symbols, this metric suggests that limiting the number of identical-symbol runs can be beneficial.

eess.SP↗

arXiv2020

Huffman-coded Sphere Shaping and Distribution Matching Algorithms via Lookup Tables

In this paper, we study amplitude shaping schemes for the probabilistic amplitude shaping (PAS) framework as well as algorithms for constant-composition distribution matching (CCDM). Huffman-coded sphere shaping (HCSS) is discussed in detail, which internally uses Huffman coding to determine the composition to be used and relies on conventional CCDM algorithms for mapping and demapping. Numerical simulations show that HCSS closes the performance gap between distribution matching schemes and sphere shaping techniques such as enumerative sphere shaping (ESS). HCSS is based on an architecture that is different from the trellis-based setup of ESS. It allows to tailor the used HCSS compositions to the transmission channel and to take into account complexity constraints. We further discuss in detail multiset ranking (MR) and subset ranking (SR) as alternatives to arithmetic-coding (AC) CCDM. The advantage of MR over AC is that it requires less sequential operations for mapping. SR operates on binary alphabets only, which can introduce some additional rate loss when a nonbinary-to-binary transformation is required. However, the binomial coefficients required for SR can be precomputed and stored in a lookup table (LUT). We perform an analysis of rate loss and decoding performance for the proposed techniques and compare them to other prominent amplitude shaping schemes. For medium to long block lengths, MR-HCSS and SR-HCSS are shown to have similar performance to ESS. SR-HCSS and uniform 64QAM are compared in additive white Gaussian noise simulations and shaping gains of 0.5 dB and 1 dB are demonstrated with 1 kbit and 100 kbit LUT size, respectively.

eess.SP↗