Springe direkt zu Inhalt

Disputation Huy Le Duc

14.07.2023 | 12:00
Thema der Dissertation:
On-premise containerized, light-weight software solutions for Biomedicine
Thema der Disputation:
Distributed Projection-based Methods for Large-Scale Linear Systems
Abstract: Large-scale linear systems pose significant challenges in terms of computational complexity, memory requirements, and the need for efficient distributed algorithms to handle the growing demands of machine learning and scientific computing applications. In this presentation, we will address these general challenges and explore potential solutions by examining the Accelerated Projection-Based  Consensus method proposed by Azizan-Ruhi et al1.
This innovative method has garnered attention due to its effectiveness in improving convergence rates for distributed computations and its unique approach to solving large-scale linear systems efficiently. One reason the work of Azizan-Ruhi et al. is particularly relevant and important is that their proposed Accelerated Projection-Based Consensus method demonstrates significant improvements in convergence rates compared to existing distributed methods. Furthermore, this acceleration can lead to substantial time savings and resource efficiency, making it a valuable addition to distributed computing.
Focusing on convergence rate analysis, we compare the Accelerated Projection-Based Consensus method with various alternative distributed methods. By investigating the mathematical foundations and performance of the Accelerated Projection-Based Consensus method in terms of convergence, we aim to highlight its potential implications and efficacy for future research and practical applications.
In conclusion, the Accelerated Projection-Based Consensus method presents a promising approach to tackling the challenges associated with large-scale linear systems. By demonstrating improved convergence rates, time savings, and resource efficiency, this method has the potential to advance the field of distributed computing significantly. Further research and practical applications of the Accelerated Projection-Based Consensus method will undoubtedly continue to uncover new insights and opportunities for enhancing the performance of distributed algorithms in machine learning and scientific computing applications.
1 N. Azizan-Ruhi, F. Lahouti, S. Avestimehr and B. Hassibi, "Distributed Solution of Large-Scale Linear Systems Via Accelerated Projection-Based Consensus," 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Calgary, AB, Canada, 2018, pp. 6358-6362, doi: 10.1109/ICASSP.2018.8462630.

Zeit & Ort

14.07.2023 | 12:00

Seminarraum 2006
(Zuse Institut Berlin, Takustr.7, 14195 Berlin)