An original parallel implementation of the local meshless numerical method in solving system of coupled partial differential equations is explored with an OpenMP based parallelization. The numerical experiments are performed on the de Vahl Davis natural convection case, with superlinear computational speedup regime identified. The phenomenon is further investigated through measurements of the cache hit rates. It is demonstrated that the accumulation of L3 caches governs the superlinear speedup.
COBISS.SI-ID: 27339815
We have proposed a new approach for online construction of classifiers based on generative models for probability density estimation. The model was derived from the online Kernel Density Estimation approach, oKDE, which we have published in Pattern Recognition journal in 2011. The new approach incrementally builds classifiers from streams of data by means of KDE. In contrast to oKDE, the new approach applies a discriminative cost function in the model simplification. Results show that the proposed approach results in more compact models than oKDE, with higher discriminative power. It yields comparable discriminative performance to batch approaches but with smaller computational complexity which allows fast construction of generative models with high discriminative power. The proposed approaches are especially suitable for robotic systems proposed in this project since they allow efficient constructions of models from streams of data at low computational complexity.
COBISS.SI-ID: 9907284
Cryotherapy has a beneficial impact on rehabilitation after injury orsurgery. However, therapeutic protocols and devices are not standardized, and clinical outcome can therefore not be related accurately to the cryotherapeutic protocols. We implement a study for Wegener, Belgium, to obtain in vivo data about temperatures of the inner knee parts after ACL reconstruction and to assess the differences between two cryotherapeutic methods: conventional cooling with gelpack, and computer-controlled cryotherapy. The cooling was performed over 12 hours. Temperatures were measured electronically every second for 16 hours after surgery in 20 patients following ACL reconstruction at the University clinical Centre Ljubljana. We have shown that the computer-controlled device outperforms the classical method of cooling with its significantly higher effectiveness and its ability to control the heat transfer. We hold an SI patent: "Method and device for noninvasive realtime control of inner temperature variables during cryo or thermo therapy" on the same topic.
COBISS.SI-ID: 26914599
The paper tackles simulation of a solidification of a binary alloy Sn-10%Pb. The problem is highly unstable due to the numerical and physical instabilities. We addressed the numerical instabilities with the techniques that alleviate the instability, however also introduce high numerical error. With the adaptive methods we minimized the negative effect. Although the problem is formulated globally we solve it with completely local approach and implemented it in parallel
COBISS.SI-ID: 3218939
The usage of multi-core networking devices itself, however, does not improve the performance, unless devices are modified in a way that all available hardware resources are utilized. Therefore, to make multi-core networking devices efficient, the process of handling the network traffic must be parallelized, which can be achieved by two different concepts: a) by a distribution of the network traffic among available processor cores and b) by a parallel implementation of offered network functionalities. We described the development and validation of an innovative improvement, based on an adaptive network-traffic-distribution on available processing cores. We confirmed theoretically and experimentally that the proposed method significantly contributes to increased performance and efficiency of multi-core networking devices.
COBISS.SI-ID: 27699495