A model is worth tens of thousands of examples - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

A model is worth tens of thousands of examples

Résumé

Traditional signal processing methods relying on mathematical data generation models have been cast aside in favour of deep neural networks, which require vast amounts of data. Since the theoretical sample complexity is nearly impossible to evaluate, these amounts of examples are usually estimated with crude rules of thumb. However, these rules only suggest when the networks should work, but do not relate to the traditional methods. In particular, an interesting question is: how much data is required for neural networks to be on par or outperform, if possible, the traditional model-based methods? In this work, we empirically investigate this question in two simple examples, where the data is generated according to precisely defined mathematical models, and where well-understood optimal or state-of-the-art mathematical data-agnostic solutions are known. A first problem is deconvolving onedimensional Gaussian signals and a second one is estimating a circle's radius and location in random grayscale images of disks. By training various networks, either naive custom designed or well-established ones, with various amounts of training data, we find that networks require tens of thousands of examples in comparison to the traditional methods, whether the networks are trained from scratch or even with transferlearning or finetuning.
Fichier principal
Vignette du fichier
230314SSVMFINALA_model_is_worth_tens_of_thousands_of_examples__SSVM_.pdf (1.6 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04271603 , version 1 (06-11-2023)

Identifiants

Citer

Thomas Dages, Laurent D. Cohen, Alfred M. Bruckstein. A model is worth tens of thousands of examples. 9th International Conference on Scale Space and Variational Methods in Computer Vision (SSVM 2023), May 2023, Santa Margherita di Pula, Italy. ⟨10.1007/978-3-031-31975-4_17⟩. ⟨hal-04271603⟩
11 Consultations
7 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More