Abstract
The growth in cloud computing, Big Data, AI and high-performance computing (HPC) necessitate the deployment of additional data centers (DC's) with high energy demands. The unprecedented increase in the Thermal Design Power (TDP) of the computing chips will require innovative cooling techniques. Furthermore, DC's are increasingly limited in their ability to add powerful GPU servers by power capacity constraints. As cooling energy use accounts for up to 40% of DC energy consumption, creative cooling solutions are urgently needed to allow deployment of additional servers, enhance sustainability and increase energy efficiency of DC's. The information in this study is provided from Start Campus' Sines facility supported by Alfa Laval for the heat exchanger and CO2 emission calculations. The study evaluates the performance and sustainability impact of various data center cooling strategies including an air-only deployment and a subsequent hybrid air/water cooling solution all utilizing sea water as the cooling source. We evaluate scenarios from 3 MW to 15+1 MW of IT load in 3 MW increments which correspond to the size of heat exchangers used in the Start Campus' modular system design. This study also evaluates the CO2 emissions compared to a conventional chiller system for all the presented scenarios. Results indicate that the effective use of the sea water cooled system combined with liquid cooled systems improve the efficiency of the DC, plays a role in decreasing the CO2 emissions and supports in achieving sustainability goals. © 2013 IEEE.

Abstract
We present an adequacy theorem for a concurrent extension of probabilistic GCL. The underlying denotational semantics is based on the so-called mixed powerdomains, which combine non-determinism with probabilistic behaviour. The theorem itself is formulated via M. Smyth's idea of treating observable properties as open sets of a topological space. The proof hinges on a 'topological generalisation' of Konig's lemma in the setting of probabilistic programming (a result that is proved in the paper as well). One application of the theorem is that it entails semi-decidability w.r.t. whether a concurrent program satisfies an observable property (written in a certain form). This is related to M. Escardo's conjecture about semi-decidability w.r.t. may and must probabilistic testing.

Abstract
Data loaders are used by Machine Learning (ML) frameworks like PyTorch and TensorFlow to apply transformations to data before feeding it into the accelerator. This operation is called data preprocessing. Data preprocessing plays an important role in the ML training workflow because if it is inefficiently pipelined with the training, it can yield high GPU idleness, resulting in important training delays. Unfortunately, existing data loaders turn out to waste GPU resources, with $76\%$ GPU idleness when using the PyTorch data loader, for example. One key source of inefficiency is the variability in preprocessing time across samples within the same dataset. Existing data loaders are oblivious to this variability, and they construct batches without any consideration of slow or fast samples. In this case, the entire batch is delayed by a single slow sample, stalling the training pipeline and resulting in head-of-line blocking. To address these inefficiencies, we present MinatoLoader, a general-purpose data loader for PyTorch that accelerates training and improves GPU utilization. MinatoLoader is designed for a single-server setup, containing multiple GPUs. It continuously prepares data in the background and actively constructs batches by prioritizing fast-to-preprocess samples, while slower samples are processed in parallel. We evaluate MinatoLoader on servers with V100 and A100 GPUs. On a machine with four A100 GPUs, MinatoLoader improves the training time of a wide range of workloads by up to $7.5\times$ ($3.6\times$ on average) over PyTorch DataLoader and Pecan, and up to $3\times$ ($2.2\times$ on average) over DALI. It also increases average GPU utilization from 46.4\% with PyTorch to 90.45\%, while preserving model accuracy and enabling faster convergence.

Abstract
To address the challenge of detecting stealthy port scans in high-speed networks, this paper introduces p4SD, a lightweight anomaly detection system that identifies reconnaissance activities directly within programmable data planes. Leveraging the P4 language, p4SD uses a cyclic fingerprint buffer and frequency analysis to monitor for anomalous traffic without relying on attack signatures. The system is designed to detect both fast and slow port scans, as its method of measuring relative changes in distinct fingerprints between cycles effectively identifies both the rapid spikes from fast scans and the gradual increases from slow scans. The proof-of-concept demonstrates resource efficiency, achieving throughput close to the hardware's theoretical limits, detecting scan activity in near real-time, and enabling timely responses to potential threats. With over 99% detection accuracy for slow scans, these findings establish p4SD as a practical and scalable solution for real-time, in-network threat detection in modern SDN environments. © 2025 University of Split, FESB.

Abstract
We present BAE, a problem-tailored and noise-aware Bayesian algorithm for quantum amplitude estimation. In a fault tolerant scenario, BAE is capable of saturating the Heisenberg limit; if device noise is present, BAE can dynamically characterize it and self-adapt. We further propose aBAE, an annealed variant of BAE drawing on methods from statistical inference, to enhance robustness. Our proposals are parallelizable in both quantum and classical components, offer tools for fast noise model assessment, and can leverage preexisting information. Additionally, they accommodate experimental limitations and preferred cost trade-offs. We propose a robust benchmark for amplitude estimation algorithms and use it to test BAE against other approaches, demonstrating its competitive performance in both noisy and noiseless scenarios. In both cases, it achieves lower error than any other algorithm as a function of the cost. In the presence of decoherence, it is capable of learning when other algorithms fail. © 2025 Elsevier B.V., All rights reserved.

Abstract
Adaptive variational quantum algorithms arguably offer the best prospects for quantum advantage in the Noisy Intermediate-Scale Quantum era. Since the inception of the first such algorithm, the Adaptive Derivative-Assembled Problem-Tailored Variational Quantum Eigensolver (ADAPT-VQE), many improvements have appeared in the literature. We combine the key improvements along with a novel operator pool-which we term Coupled Exchange Operator (CEO) pool-to assess the cost of running state-of-the-art ADAPT-VQE on hardware in terms of measurement counts and circuit depth. We show a dramatic reduction of these quantum computational resources compared to the early versions of the algorithm: CNOT count, CNOT depth and measurement costs are reduced by up to 88%, 96% and 99.6%, respectively, for molecules represented by 12 to 14 qubits (LiH, H6 and BeH2). We also find that our state-of-the-art CEO-ADAPT-VQE outperforms the Unitary Coupled Cluster Singles and Doubles ansatz, the most widely used static VQE ansatz, in all relevant metrics, and offers a five order of magnitude decrease in measurement costs as compared to other static ans & auml;tze with competitive CNOT counts.