AdaSwarm: Augmenting Gradient-Based optimizers in Deep Learning with Swarm Intelligence

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Mohapatra, Rohan, Saha, Snehanshu, Coello, Carlos A. Coello, Bhattacharya, Anwesh, Dhavala, Soma S., Saha, Sriparna
Formato: Preprint
Publicado: 2020
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911887513354240
author Mohapatra, Rohan
Saha, Snehanshu
Coello, Carlos A. Coello
Bhattacharya, Anwesh
Dhavala, Soma S.
Saha, Sriparna
author_facet Mohapatra, Rohan
Saha, Snehanshu
Coello, Carlos A. Coello
Bhattacharya, Anwesh
Dhavala, Soma S.
Saha, Sriparna
contents This paper introduces AdaSwarm, a novel gradient-free optimizer which has similar or even better performance than the Adam optimizer adopted in neural networks. In order to support our proposed AdaSwarm, a novel Exponentially weighted Momentum Particle Swarm Optimizer (EMPSO), is proposed. The ability of AdaSwarm to tackle optimization problems is attributed to its capability to perform good gradient approximations. We show that, the gradient of any function, differentiable or not, can be approximated by using the parameters of EMPSO. This is a novel technique to simulate GD which lies at the boundary between numerical methods and swarm intelligence. Mathematical proofs of the gradient approximation produced are also provided. AdaSwarm competes closely with several state-of-the-art (SOTA) optimizers. We also show that AdaSwarm is able to handle a variety of loss functions during backpropagation, including the maximum absolute error (MAE).
format Preprint
id arxiv_https___arxiv_org_abs_2006_09875
institution arXiv
publishDate 2020
record_format arxiv
spellingShingle AdaSwarm: Augmenting Gradient-Based optimizers in Deep Learning with Swarm Intelligence
Mohapatra, Rohan
Saha, Snehanshu
Coello, Carlos A. Coello
Bhattacharya, Anwesh
Dhavala, Soma S.
Saha, Sriparna
Neural and Evolutionary Computing
Machine Learning
This paper introduces AdaSwarm, a novel gradient-free optimizer which has similar or even better performance than the Adam optimizer adopted in neural networks. In order to support our proposed AdaSwarm, a novel Exponentially weighted Momentum Particle Swarm Optimizer (EMPSO), is proposed. The ability of AdaSwarm to tackle optimization problems is attributed to its capability to perform good gradient approximations. We show that, the gradient of any function, differentiable or not, can be approximated by using the parameters of EMPSO. This is a novel technique to simulate GD which lies at the boundary between numerical methods and swarm intelligence. Mathematical proofs of the gradient approximation produced are also provided. AdaSwarm competes closely with several state-of-the-art (SOTA) optimizers. We also show that AdaSwarm is able to handle a variety of loss functions during backpropagation, including the maximum absolute error (MAE).
title AdaSwarm: Augmenting Gradient-Based optimizers in Deep Learning with Swarm Intelligence
topic Neural and Evolutionary Computing
Machine Learning
url https://arxiv.org/abs/2006.09875