Výsledky vyhledávání - "Schain, Mariano"

Report

Locally Optimal Descent for Dynamic Stepsize Scheduling

Autor: Yehudai, Gilad, Cohen, Alon, Daniely, Amit, Drori, Yoel, Koren, Tomer, Schain, Mariano

We introduce a novel dynamic learning-rate scheduling scheme grounded in theory with the goal of simplifying the manual and time-consuming tuning of schedules in practice. Our approach is based on estimating the locally-optimal stepsize, guaranteeing

Externí odkaz: http://arxiv.org/abs/2311.13877

Zobrazit plný text záznamu

Report

The Temporal Structure of Language Processing in the Human Brain Corresponds to The Layered Hierarchy of Deep Language Models

Deep Language Models (DLMs) provide a novel computational paradigm for understanding the mechanisms of natural language processing in the human brain. Unlike traditional psycholinguistic models, DLMs use layered sequences of continuous numerical vect

Externí odkaz: http://arxiv.org/abs/2310.07106

Zobrazit plný text záznamu

Report

Adversarial Robustness of Streaming Algorithms through Importance Sampling

Autor: Braverman, Vladimir, Hassidim, Avinatan, Matias, Yossi, Schain, Mariano, Silwal, Sandeep, Zhou, Samson

In this paper, we introduce adversarially robust streaming algorithms for central machine learning and algorithmic tasks, such as regression and clustering, as well as their more general counterparts, subspace embedding, low-rank approximation, and c

Externí odkaz: http://arxiv.org/abs/2106.14952

Zobrazit plný text záznamu

Report

Asynchronous Stochastic Optimization Robust to Arbitrary Delays

Autor: Cohen, Alon, Daniely, Amit, Drori, Yoel, Koren, Tomer, Schain, Mariano

We consider stochastic optimization with delayed gradients where, at each time step $t$, the algorithm makes an update using a stale stochastic gradient from step $t - d_t$ for some arbitrary delay $d_t$. This setting abstracts asynchronous distribut

Externí odkaz: http://arxiv.org/abs/2106.11879

Zobrazit plný text záznamu

Akademický článek

Tento výsledek nelze pro nepřihlášené uživatele zobrazit.
K zobrazení výsledku je třeba se přihlásit.

Report

Scalable Learning of Non-Decomposable Objectives

Autor: Eban, Elad ET., Schain, Mariano, Mackey, Alan, Gordon, Ariel, Saurous, Rif A., Elidan, Gal

Modern retrieval systems are often driven by an underlying machine learning model. The goal of such systems is to identify and possibly rank the few most relevant items for a given query or context. Thus, such systems are typically evaluated using a

Externí odkaz: http://arxiv.org/abs/1608.04802

Zobrazit plný text záznamu