Zobrazeno 1 - 10
of 4 775
pro vyhledávání: '"A. DeFazio"'
Vision language models (VLMs) can simultaneously reason about images and texts to tackle many tasks, from visual question answering to image captioning. This paper focuses on map parsing, a novel task that is unexplored within the VLM context and par
Externí odkaz:
http://arxiv.org/abs/2409.12842
Autor:
Defazio, Aaron, Yang, Xingyu Alice, Mehta, Harsh, Mishchenko, Konstantin, Khaled, Ahmed, Cutkosky, Ashok
Existing learning rate schedules that do not require specification of the optimization stopping step T are greatly out-performed by learning rate schedules that depend on T. We propose an approach that avoids the need for this stopping time by eschew
Externí odkaz:
http://arxiv.org/abs/2405.15682
We develop new sub-optimality bounds for gradient descent (GD) that depend on the conditioning of the objective along the path of optimization, rather than on global, worst-case constants. Key to our proofs is directional smoothness, a measure of gra
Externí odkaz:
http://arxiv.org/abs/2403.04081
Learning rate schedules used in practice bear little resemblance to those recommended by theory. We close much of this theory/practice gap, and as a consequence are able to derive new problem-adaptive learning rate schedules. Our main technical contr
Externí odkaz:
http://arxiv.org/abs/2310.07831
Seeing-eye robots are very useful tools for guiding visually impaired people, potentially producing a huge societal impact given the low availability and high cost of real guide dogs. Although a few seeing-eye robot systems have already been demonstr
Externí odkaz:
http://arxiv.org/abs/2309.04370
Autor:
Mishchenko, Konstantin, Defazio, Aaron
We consider the problem of estimating the learning rate in adaptive methods, such as AdaGrad and Adam. We propose Prodigy, an algorithm that provably estimates the distance to the solution $D$, which is needed to set the learning rate optimally. At i
Externí odkaz:
http://arxiv.org/abs/2306.06101
We introduce a technique for tuning the learning rate scale factor of any base optimization algorithm and schedule automatically, which we call \textsc{mechanic}. Our method provides a practical realization of recent theoretical reductions for accomp
Externí odkaz:
http://arxiv.org/abs/2306.00144
Training a modern machine learning architecture on a new task requires extensive learning-rate tuning, which comes at a high computational cost. Here we develop new Polyak-type adaptive learning rates that can be used on top of any momentum method, a
Externí odkaz:
http://arxiv.org/abs/2305.07583
Autor:
Ksenija Nesic, John J. Krais, Yifan Wang, Cassandra J. Vandenberg, Pooja Patel, Kathy Q. Cai, Tanya Kwan, Elizabeth Lieschke, Gwo-Yaw Ho, Holly E. Barker, Justin Bedo, Silvia Casadei, Andrew Farrell, Marc Radke, Kristy Shield-Artin, Jocelyn S. Penington, Franziska Geissler, Elizabeth Kyran, Robert Betsch, Lijun Xu, Fan Zhang, Alexander Dobrovic, Inger Olesen, Rebecca Kristeleit, Amit Oza, Iain McNeish, Gayanie Ratnayake, Nadia Traficante, Australian Ovarian Cancer Study, Anna DeFazio, David D. L. Bowtell, Thomas C. Harding, Kevin Lin, Elizabeth M. Swisher, Olga Kondrashova, Clare L. Scott, Neil Johnson, Matthew J. Wakefield
Publikováno v:
Molecular Cancer, Vol 23, Iss 1, Pp 1-10 (2024)
Abstract PARP inhibitor (PARPi) therapy has transformed outcomes for patients with homologous recombination DNA repair (HRR) deficient ovarian cancers, for example those with BRCA1 or BRCA2 gene defects. Unfortunately, PARPi resistance is common. Mul
Externí odkaz:
https://doaj.org/article/8f0a9b712db94dd79db01d43d570add9
Autor:
Defazio, Aaron, Mishchenko, Konstantin
D-Adaptation is an approach to automatically setting the learning rate which asymptotically achieves the optimal rate of convergence for minimizing convex Lipschitz functions, with no back-tracking or line searches, and no additional function value o
Externí odkaz:
http://arxiv.org/abs/2301.07733