arXiv subjects

Vladimir Stankovic

Publications and source records attributed to Vladimir Stankovic.

11 recordsLinked to original sources

arXiv2026

Effectiveness of LLM-based Software Diversity for Reliability Improvement -- an Empirical Study

Software diversity has been extensively studied as a means of reducing the risk of common-mode failures. Classic work showed that the central issue is whether failures of diversely redundant components overlap in ways that limit the reliability gains. Traditional software diversity is costly to obtain, since it requires multiple implementations as well as the corresponding validation, maintenance, and deployment effort. Recent advances in Large Language Models (LLMs) may change this. LLMs enable inexpensive code generation: they produce many candidate implementations of the same specification quickly, across different models, decoding settings, and programming languages. This raises a natural question: can LLMs serve as practical generators of software diversity, and how much reliability improvement can that diversity actually provide? In this paper, we extend classical empirical studies of software diversity in human-written programs to LLM-generated code. We study three specifications using both historical human-written programs and large pools of LLM-generated ones evaluated under a common compilation, sandboxing, and exhaustive test suite. We explore LLM diversity along multiple axes, including model family, generation temperature, and programming language. Reliability improvement is evaluated in a 1-out-of-2 configuration across both homogeneous and heterogeneous program populations, including within-LLM pairings and pairings across programming languages and across LLM-generated and human-written programs. The results show that combining LLM-generated programs, especially in heterogeneous settings, can yield reliability gains, although this is partly conditioned by the programming language and generation setting. Taken together, these findings suggest that LLMs provide a scalable source of comparatively low-cost programs whose diversity can be leveraged for reliability improvement.

cs.SE↗

arXiv2025

Time and Money Matters for Sustainability: Insights on User Preferences on Renewable Energy for Electric Vehicle Charging Stations

Charging electric vehicles (EVs) with renewable energy can lessen their environmental impact. However, the fluctuating availability of renewable energy affects the sustainability of public EV charging stations. Nearby public charging stations may utilize differing energy sources due to their microgrid connections - ranging from exclusively renewable to non-renewable or a combination of both - highlighting the substantial variability in energy supply types within short distances. This study investigates the near-future scenario of integrating dynamic renewable energy availability in charging station navigation to impact the choices of EV users towards renewable sources. We conducted a within-subjects design survey with 50 car users and semi-structured interviews with 10 EV users from rural, suburban, and urban areas. The results show that when choosing EV charging stations, drivers often prioritize either time savings or money savings based on the driving scenarios that influence drivers' consumer value. Notably, EV users tend to select renewable-powered stations when they align with their main priority, be it saving money or time. This study offers end-user insights into the front-end graphic user interface and the development of the back-end ranking algorithm for navigation recommender systems that integrate dynamic renewable energy availability for the sustainable use of electric vehicles.

cs.HC↗

arXiv2025

PANDAS: Peer-to-peer, Adaptive Networking for Data Availability Sampling within Ethereum Consensus Timebounds

Layer-2 protocols can assist Ethereum's limited throughput, but globally broadcasting layer-2 data limits their scalability. The Danksharding evolution of Ethereum aims to support the selective distribution of layer-2 data, whose availability in the network is verified using randomized data availability sampling (DAS). Integrating DAS into Ethereum's consensus process is challenging, as pieces of layer-2 data must be disseminated and sampled within four seconds of the beginning of each consensus slot. No existing solution can support dissemination and sampling under such strict time bounds. We propose PANDAS, a practical approach to integrate DAS with Ethereum under Danksharding's requirements without modifying its protocols for consensus and node discovery. PANDAS disseminates layer-2 data and samples its availability using lightweight, direct exchanges. Its design accounts for message loss, node failures, and unresponsive participants while anticipating the need to scale out the Ethereum network. Our evaluation of PANDAS's prototype in a 1,000-node cluster and simulations for up to 20,000 peers shows that it allows layer-2 data dissemination and sampling under planetary-scale latencies within the 4-second deadline.

cs.DC↗

arXiv2024

Classifying bioacoustic data without individual call annotations using temporal convolutional networks and feature extractors

Bioacoustic data from Passive Acoustic Monitoring (PAM) generates large datasets where obtaining detailed auditing and labelling is often impractical, resulting in weak annotations (e.g., presence/absence of species over several minutes of recording). In order to effectively capture the complex temporal patterns and key features of long audio segments, we propose a framework comprising dataset standardisation, feature extraction, and classification via Temporal Convolutional Networks (TCN). This approach eliminates the necessity for setting heuristic decision rules or creating time-consuming strong labels. To demonstrate the effectiveness of our approach, we use sperm whale (\textit{Physeter macrocephalus}) click trains in 4-minute recordings as a case study, from a dataset comprising diverse sources and deployment conditions to maximise generalisability. Our TCN classifiers achieve recall rates exceeding 0.83 at a 0.13 false positive rate, comparable to agreement rates between expert annotators. We compare two methods of feature extraction, Variational AutoEncoders (VAEs) and traditional handpicking of features, and found them to yield similar performance results, with the VAE-based classifiers seeing a more stable performance across datasets and recording conditions. These results offer a way forward in leveraging numerous existing annotated bioacoustic datasets to train automatic classification models, effectively overcoming previous limitations associated with weak labels.

cs.SD↗

arXiv2023

A Survey on Experimental Performance Evaluation of Data Distribution Service (DDS) Implementations

The Data Distribution Service (DDS) is a widely used communication specification for real-time mission-critical systems that follow the principles of publish-subscribe middleware. DDS has an extensive set of quality of service (QoS) parameters allowing a thorough customisation of the intended communication. An extensive survey of the performance of the implementations of this communication middleware is lacking. This paper closes the gap by surveying the state of the art in performance of various DDS implementations and identifying any research gaps that exist within this domain.

cs.DC↗

arXiv2023

Data Availability Sampling in Ethereum: Analysis of P2P Networking Requirements

Despite their increasing popularity, blockchains still suffer from severe scalability limitations. Recently, Ethereum proposed a novel approach to block validation based on Data Availability Sampling (DAS), that has the potential to improve its transaction per second rate by more than two orders of magnitude. DAS should also significantly reduce per-transaction validation costs. At the same time, DAS introduces new communication patterns in the Ethereum Peer-to-Peer (P2P) network. These drastically increase the amount of exchanged data and impose stringent latency objectives. In this paper, we review the new requirements for P2P networking associated with DAS, discuss open challenges, and identify new research directions.

cs.NI↗

arXiv2022

Exploring the Effects of Multicast Communication on DDS Performance

The Data Distribution Service (DDS) is an Object Management Group (OMG) standard for high-performance and real-time systems. DDS is a data-centric middleware based on the publish-subscribe communication pattern and is used in many mission-critical, or even safety-critical, systems such as air traffic control and robot operating system (ROS2). This research aims at identifying how the usage of multicast affects the performance of DDS communication for varying numbers of participants (publishers and subscribers). The results show that DDS configured for multicast communication can exhibit worse performance under a high load (a greater number of participants) than DDS configured for unicast communication. This counter-intuitive result reinforces the need for researchers and practitioners to be clear about the details of how multicast communication operates on the network.

cs.PF↗

arXiv2021

How likely is a random graph shift-enabled?

The shift-enabled property of an underlying graph is essential in designing distributed filters. This article discusses when a random graph is shift-enabled. In particular, popular graph models ER, WS, BA random graph are used, weighted and unweighted, as well as signed graphs. Our results show that the considered unweighted connected random graphs are shift-enabled with high probability when the number of edges is moderately high. However, very dense graphs, as well as fully connected graphs, are not shift-enabled. Interestingly, this behaviour is not observed for weighted connected graphs, which are always shift-enabled unless the number of edges in the graph is very low.

cs.DM↗

arXiv2019

Robust Deep Graph Based Learning for Binary Classification

Convolutional neural network (CNN)-based feature learning has become state of the art, since given sufficient training data, CNN can significantly outperform traditional methods for various classification tasks. However, feature learning becomes more difficult if some training labels are noisy. With traditional regularization techniques, CNN often overfits to the noisy training labels, resulting in sub-par classification performance. In this paper, we propose a robust binary classifier, based on CNNs, to learn deep metric functions, which are then used to construct an optimal underlying graph structure used to clean noisy labels via graph Laplacian regularization (GLR). GLR is posed as a convex maximum a posteriori (MAP) problem solved via convex quadratic programming (QP). To penalize samples around the decision boundary, we propose two regularized loss functions for semi-supervised learning. The binary classification experiments on three datasets, varying in number and type of features, demonstrate that given a noisy training dataset, our proposed networks outperform several state-of-the-art classifiers, including label-noise robust support vector machine, CNNs with three different robust loss functions, model-based GLR, and dynamic graph CNN classifiers.

cs.LG↗

arXiv2019

Undirected graphs: is the shift-enabled condition trivial or necessary?

It has recently been shown that, contrary to the wide belief that a shift-enabled condition (necessary for any shift-invariant filter to be representable by a graph shift matrix) can be ignored because any non-shift-enabled matrix can be converted to a shift-enabled matrix, such a conversion in general may not hold for a directed graph with non-symmetric shift matrix. This letter extends this prior work, focusing on undirected graphs where the shift matrix is generally symmetric. We show that while, in this case, the shift matrix can be converted to satisfy the original shift-enabled condition, the converted matrix is not associated with the original graph, that is, it does not capture anymore the structure of the graph signal. We show via a counterexample, that a non-shift-enabled matrix cannot be converted to a shift-enabled one and still maintain the topological structure of the underlying graph, which is necessary to facilitate localized signal processing.

eess.SP↗

arXiv2018

Linear State Estimation via 5G C-RAN Cellular Networks using Gaussian Belief Propagation

Machine-type communications and large-scale information processing architectures are among key (r)evolutionary enhancements of emerging fifth-generation (5G) mobile cellular networks. Massive data acquisition and processing will make 5G network an ideal platform for large-scale system monitoring and control with applications in future smart transportation, connected industry, power grids, etc. In this work, we investigate a capability of such a 5G network architecture to provide the state estimate of an underlying linear system from the input obtained via large-scale deployment of measurement devices. Assuming that the measurements are communicated via densely deployed cloud radio access network (C-RAN), we formulate and solve the problem of estimating the system state from the set of signals collected at C-RAN base stations. Our solution, based on the Gaussian Belief-Propagation (GBP) framework, allows for large-scale and distributed deployment within the emerging 5G information processing architectures. The presented numerical study demonstrates the accuracy, convergence behavior and scalability of the proposed GBP-based solution to the large-scale state estimation problem.

cs.IT↗