1. High-performance deployment operational Data analytics of pre-trained multi-label classification architectures with differential-evolution-based hyperparameter optimization (AutoDEHypO)Teo Prica, Aleš Zamuda, 2025, original scientific article Abstract: This article presents a high-performance-computing differential-evolution-based hyperparameter optimization automated workflow (AutoDEHypO), which is deployed on a petascale supercomputer and utilizes multiple GPUs to execute a specialized fitness function for machine learning (ML). The workflow is designed for operational analytics of energy efficiency. In this differential evolution (DE) optimization use case, we analyze how energy efficiently the DE algorithm performs with different DE strategies and ML models. The workflow analysis considers key factors such as DE strategies and automated use case configurations, such as an ML model architecture and dataset, while monitoring both the achieved accuracy and the utilization of computing resources, such as the elapsed time and consumed energy. While the efficiency of a chosen DE strategy is assessed based on a multi-label supervised ML accuracy, operational data about the consumption of resources of individual completed jobs obtained from a Slurm database are reported. To demonstrate the impact on energy efficiency, using our analysis workflow, we visualize the obtained operational data and aggregate them with statistical tests that compare and group the energy efficiency of the DE strategies applied in the ML models.
Keywords: high-performance computing, operational data analytics, energy efficiency, machine learning, AutoML, differential avolution, optimization
Published in DKUM: 29.05.2025; Views: 0; Downloads: 9
 Full text (1,61 MB) |
2. Parallel self-avoiding walks for a low-autocorrelation binary sequences problemBorko Bošković, Jana Herzog, Janez Brest, 2024, original scientific article Abstract: A low-autocorrelation binary sequences problem with a high figure of merit factor represents a formidable computational challenge. An efficient parallel computing algorithm is required to reach the new best-known solutions for this problem. Therefore, we developed the sokol solver for the skew-symmetric search space. The developed solver takes the advantage of parallel computing on graphics processing units. The solver organized the search process as a sequence of parallel and contiguous self-avoiding walks and achieved a speedup factor of 387 compared with lssOrel, its predecessor. The sokol solver belongs to stochastic solvers and cannot guarantee the optimality of solutions. To mitigate this problem, we established the predictive model of stopping conditions according to the small instances for which the optimal skew-symmetric solutions are known. With its help and 99% probability, the sokol solver found all the known and seven new best-known skew-symmetric sequences for odd instances from to . For larger instances, the solver cannot reach 99% probability within our limitations, but it still found several new best-known binary sequences. We also analyzed the trend of the best merit factor values, and it shows that as sequence size increases, the value of the merit factor also increases, and this trend is flatter for larger instances.
Keywords: low-autocorrelation binary sequences, self-avoiding walk, graphic processor units, high performance computing
Published in DKUM: 22.08.2024; Views: 45; Downloads: 23
Full text (1,82 MB)
This document has many files! More... |
3. NiaNetAD: Autoencoder architecture search for tabular anomaly detection powered by HPCSašo Pavlič, Sašo Karakatič, Iztok Fister, 2023, published scientific conference contribution Keywords: predictive maintenance, autoencoder, anomaly detection, nature-inspired algorithms, optimization, high-performance computing, unsupervised learning
Published in DKUM: 26.01.2024; Views: 447; Downloads: 0 |
4. |
