مؤتمر
A new non-convex framework to improve asymptotical knowledge on generic stochastic gradient descent
العنوان: | A new non-convex framework to improve asymptotical knowledge on generic stochastic gradient descent |
---|---|
المؤلفون: | Fest, Jean-Baptiste, Repetti, Audrey, Chouzenoux, Emilie |
المساهمون: | OPtimisation Imagerie et Santé (OPIS), Inria Saclay - Ile de France, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre de vision numérique (CVN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-CentraleSupélec-Université Paris-Saclay, Centre de vision numérique (CVN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay, Heriot-Watt University Edinburgh (HWU), European Project: ERC-2019-STG-850925,MAJORIS(2020) |
المصدر: | Proceedings of the IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2023) ; MLSP 2023 - IEEE International Workshop on Machine Learning for Signal Processing ; https://inria.hal.science/hal-04165342Test ; MLSP 2023 - IEEE International Workshop on Machine Learning for Signal Processing, Sep 2023, Rome, Italy |
بيانات النشر: | HAL CCSD |
سنة النشر: | 2023 |
المجموعة: | Archive ouverte HAL (Hyper Article en Ligne, CCSD - Centre pour la Communication Scientifique Directe) |
مصطلحات موضوعية: | Stochastic gradient descent, non-convex optimization, Kurdyka-Lojasiewicz, convergence analysis, [MATH.MATH-OC]Mathematics [math]/Optimization and Control [math.OC] |
جغرافية الموضوع: | Rome, Italy |
الوصف: | International audience ; Stochastic gradient optimization methods are broadly used to minimize non-convex smooth objective functions, for instance when training deep neural networks. However, theoretical guarantees on the asymptotic behaviour of these methods remain scarce. Especially, ensuring almost-sure convergence of the iterates to a stationary point is quite challenging. In this work, we introduce a new Kurdyka-Łojasiewicz theoretical framework to analyze asymptotic behavior of stochastic gradient descent (SGD) schemes when minimizing non-convex smooth objectives. In particular, our framework provides new almost-sure convergence results, on iterates generated by any SGD method satisfying mild conditional descent conditions. We illustrate the proposed framework by means of several toy simulation examples. We illustrate the role of the considered theoretical assumptions, and investigate how SGD iterates are impacted whether these assumptions are either fully or partially satisfied. |
نوع الوثيقة: | conference object |
اللغة: | English |
العلاقة: | info:eu-repo/grantAgreement//ERC-2019-STG-850925/EU/Majoration-Minimization algorithms for Image Processing/MAJORIS; hal-04165342; https://inria.hal.science/hal-04165342Test; https://inria.hal.science/hal-04165342/documentTest; https://inria.hal.science/hal-04165342/file/MLSP_2023.pdfTest |
الإتاحة: | https://inria.hal.science/hal-04165342Test https://inria.hal.science/hal-04165342/documentTest https://inria.hal.science/hal-04165342/file/MLSP_2023.pdfTest |
حقوق: | http://creativecommons.org/licenses/byTest/ ; info:eu-repo/semantics/OpenAccess |
رقم الانضمام: | edsbas.4431CF23 |
قاعدة البيانات: | BASE |
الوصف غير متاح. |