Zobrazeno 1 - 10
of 123
pro vyhledávání: '"Idreos, Stratos"'
While transformers have been at the core of most recent advancements in sequence generative models, their computational cost remains quadratic in sequence length. Several subquadratic architectures have been proposed to address this computational iss
Externí odkaz:
http://arxiv.org/abs/2410.12982
Autor:
Liang, Wanchao, Liu, Tianyu, Wright, Less, Constable, Will, Gu, Andrew, Huang, Chien-Chin, Zhang, Iris, Feng, Wei, Huang, Howard, Wang, Junjie, Purandare, Sanket, Nadathur, Gokul, Idreos, Stratos
The development of large language models (LLMs) has been instrumental in advancing state-of-the-art natural language processing applications. Training LLMs with billions of parameters and trillions of tokens require sophisticated distributed systems
Externí odkaz:
http://arxiv.org/abs/2410.06511
Memory disaggregation (MD) allows for scalable and elastic data center design by separating compute (CPU) from memory. With MD, compute and memory are no longer coupled into the same server box. Instead, they are connected to each other via ultra-fas
Externí odkaz:
http://arxiv.org/abs/2207.03027
Autor:
Knorr, Eric R., Lemaire, Baptiste, Lim, Andrew, Luo, Siqiang, Zhang, Huanchen, Idreos, Stratos, Mitzenmacher, Michael
Publikováno v:
SIGMOD'22: Proceedings of the 2022 International Conference on Management of Data, June 2022, 1670-1684
We introduce Proteus, a novel self-designing approximate range filter, which configures itself based on sampled data in order to optimize its false positive rate (FPR) for a given space requirement. Proteus unifies the probabilistic and deterministic
Externí odkaz:
http://arxiv.org/abs/2207.01503
Real-time analytics systems employ hybrid data layouts in which data are stored in different formats throughout their lifecycle. Recent data are stored in a row-oriented format to serve OLTP workloads and support high insert rates, while older data a
Externí odkaz:
http://arxiv.org/abs/2101.06801
Autor:
Idreos, Stratos, Dayan, Niv, Qin, Wilson, Akmanalp, Mali, Hilgard, Sophie, Ross, Andrew, Lennon, James, Jain, Varun, Gupta, Harshita, Li, David, Zhu, Zichen
We introduce the concept of design continuums for the data layout of key-value stores. A design continuum unifies major distinct data structure designs under the same model. The critical insight and potential long-term impact is that such unifying mo
Externí odkaz:
http://arxiv.org/abs/1907.05443
Akademický článek
Tento výsledek nelze pro nepřihlášené uživatele zobrazit.
K zobrazení výsledku je třeba se přihlásit.
K zobrazení výsledku je třeba se přihlásit.
Ensembles of deep neural networks significantly improve generalization accuracy. However, training neural network ensembles requires a large amount of computational resources and time. State-of-the-art approaches either train all networks from scratc
Externí odkaz:
http://arxiv.org/abs/1809.04270
Data structures are critical in any data-driven scenario, but they are notoriously hard to design due to a massive design space and the dependence of performance on workload and hardware which evolve continuously. We present a design engine, the Data
Externí odkaz:
http://arxiv.org/abs/1808.02066
Anyone in need of a data system today is confronted with numerous complex options in terms of system architectures, such as traditional relational databases, NoSQL and NewSQL solutions as well as several sub-categories like column-stores, row-stores
Externí odkaz:
http://arxiv.org/abs/1706.05714