Deep neural networks (DNNs), particularly convolutional neural networks (CNNs), have garnered significant attention in recent years for addressing a wide range of challenges in image processing and computer vision. Neural architecture search (NAS) has emerged as a crucial field aiming to automate the design and configuration of CNN models. In this paper, we propose a novel strategy to speed up the performance estimation of neural architectures by gradually increasing the size of the training set used for evaluation as the search progresses. We evaluate this approach using the CGP-NASV2 model, a multi-objective NAS method, on the CIFAR-100 dataset. Experimental results demonstrate a notable acceleration in the search process, achieving a speedup of 4.6 times compared to the baseline. Despite using limited data in the early stages, our proposed method effectively guides the search towards competitive architectures. This study highlights the efficacy of leveraging lower-fidelity estimates in NAS and paves the way for further research into accelerating the design of efficient CNN architectures.
@article{Garcia-Garcia2024B,author={Garcia-Garcia, Cosijopii and Derbel, Bilel and Morales-Reyes, Alicia and Escalante, Hugo Jair},editor={Mart{\'i}nez-Villase{\~{n}}or, Lourdes and Ochoa-Ruiz, Gilberto},title={Speeding up the Multi-objective NAS Through Incremental Learning},booktitle={Advances in Soft Computing},journal={23rd Mexican International Conference on Artificial Intelligence},year={2025},publisher={Springer Nature Switzerland},address={Cham},pages={3--15},isbn={978-3-031-75543-9}}
2024
EvoStar
Progressive Self-supervised Multi-objective NAS for Image Classification
Cosijopii Garcia-Garcia, Alicia Morales-Reyes, and Hugo Jair Escalante
International Conference on the Applications of Evolutionary Computation (Part of EvoStar), 2024
We introduce a novel progressive self-supervised framework for neural architecture search. Our aim is to search for competitive, yet significantly less complex, generic CNN architectures that can be used for multiple tasks (i.e., as a pretrained model). This is achieved through cartesian genetic programming (CGP) for neural architecture search (NAS). Our approach integrates self-supervised learning with a progres- sive architecture search process. This synergy unfolds within the continu- ous domain which is tackled via multi-objective evolutionary algorithms (MOEAs). To empirically validate our proposal, we adopted a rigorous evaluation using the non-dominated sorting genetic algorithm II (NSGA- II) for the CIFAR-100, CIFAR-10, SVHN and CINIC-10 datasets. The experimental results showcase the competitiveness of our approach in relation to state-of-the-art proposals concerning both classification per- formance and model complexity. Additionally, the effectiveness of this method in achieving strong generalization can be inferred.
@article{Garcia-Garcia2024A,author={Garcia-Garcia, Cosijopii and Morales-Reyes, Alicia and Escalante, Hugo Jair},doi={10.1007/978-3-031-56855-8_11},pages={180--195},title={{Progressive Self-supervised Multi-objective NAS for Image Classification}},year={2024},journal={International Conference on the Applications of Evolutionary Computation (Part of EvoStar)}}
2023
ASOC
Continuous Cartesian Genetic Programming based representation for multi-objective neural architecture search
Cosijopii Garcia-Garcia, Alicia Morales-Reyes, and Hugo Jair Escalante
We propose a novel neural architecture search (NAS) approach for the challenge of designing convolutional neural networks (CNNs) that achieve a good tradeoff between complexity and accuracy. We rely on Cartesian genetic programming (CGP) and integrated real-based and block-chained CNN representation, for optimization using multi-objective evolutionary algorithms (MOEAs) in the continuous domain. We introduce two variants, CGP-NASV1 and CGP-NASV2, which differ in the granularity of their respective search spaces. To evaluate the proposed algorithms, we utilized the non-dominated sorting genetic algorithm II (NSGA-II) on the CIFAR-10, CIFAR-100,and SVHN datasets. Additionally, we extended the empirical analysis while maintaining the same solution representation to assess other searching techniques such as differential evolution (DE), the multi-objective evolutionary algorithm based on decomposition (MOEA/D), and the S metric selection evolutionary multi-objective algorithm (SMS-EMOA). The experimental results demonstrate that our approach exhibits competitive classification performance and model complexity compared to state-of-the-art methods.
@article{Garcia-Garcia2023,title={Continuous Cartesian Genetic Programming based representation for multi-objective neural architecture search},journal={Applied Soft Computing},volume={147},pages={110788},year={2023},issn={1568-4946},month=sep,doi={10.1016/j.asoc.2023.110788},author={Garcia-Garcia, Cosijopii and Morales-Reyes, Alicia and Escalante, Hugo Jair},keywords={Neural architecture search, Cartesian genetic programming, Convolutional neural network, Multi-objective optimization}}
2022
Tech Report
Automatic design of convolutional neural
network architectures using Multiobjective
Evolutionary Algorithms
Convolutional Neural Networks (CNNs) have had a remarkable performance in difficult computer vision tasks. In previous years, human experts have developed a number of specialized CNN archi-tectures to deal with complex image datasets. However, the automatic design of CNN through Neural Architecture Search (NAS) has gained importance to reduce and possibly avoid human expert intervention. One of the main challenges in NAS is to design less complex and yet highly precise CNNs when both objectives conflict. This study extends Cartesian Genetic Programming (CGP) for CNNs representation in NAS through multi-objective evolutionary optimization for image classification tasks. The proposed CGP-NAS algorithm is built on CGP by combining real-based solutions representation and the well-established Non-dominated Sorting Genetic Algorithm II (NSGA-II). A detailed empirical assessment shows CGP-NAS achieved competitive performance when compared to other state-of-the-art proposals while significantly reduced the evolved CNNs architecture’s complexity as well as GPU-days.
@article{Garcia-Garcia2022,author={Garcia-Garcia, Cosijopii and Escalante, Hugo Jair and Morales-Reyes, Alicia},doi={10.1145/3520304.3528963},isbn={9781450392686},journal={Genetic and Evolutionary Computation Conference Companion (GECCO '22 Companion), July 9–13, 2022, Boston, MA, USA},keywords={CGP,CNN,Neural architecture search,image classification,multi-objective evolutionary optimization},title={{CGP-NAS: Real-based solutions encoding for multi-objective evolutionary neural architecture search}},volume={1},year={2022},month=jul}
2020
GECCO
cMOGA/D: a novel cellular GA based on decomposition to tackle constrained multiobjetive problems
Cosijopii Garcia-Garcia, María-Guadalupe Martínez-Peñaloza, and Alicia Morales-Reyes
In Genetic and Evolutionary Computation Conference Companion (GECCO ’20 Companion),, Jul 2020
Parallel schemes in Evolutionary Algorithms (EAs) and in particular fine-grained or cellular Genetic Algorithms (cGAs) to tackle constrained multiobjective optimization problems (CMOPs) have been narrowly investigated. Cellular GAs structural properties balance exploration and exploitation stages during the search to avoid stagnation and possible premature convergence. This article proposes cMOGA/D, a novel algorithmic approach that takes advantage of structural properties in cGAs in combination with core concepts of well established MOEA/D technique, such as, targeting multiob-jective problems via multiple subproblems by Tchebycheff decomposition ; its neighbouring conceptual basis; and also combining a CMOP Constraint Handling Technique (CHT) called push and pull search (PPS). A thorough empirical assessment shows an improved cMOGA/D performance in comparison to previous state of the art approaches.
@article{Garcia-Garcia2020,address={Canc{\'{u}}n, Mexico},author={Garcia-Garcia, Cosijopii and Mart{\'{i}}nez-Pe{\~{n}}aloza, Mar{\'{i}}a-Guadalupe and Morales-Reyes, Alicia},journal={In Genetic and Evolutionary Computation Conference Companion (GECCO '20 Companion),},doi={10.1145/3377929.3398137},isbn={9781450371278},keywords={Cellular genetic algorithms,Constrained multiobjective optimization problems,Multiojective optimization,Push and Pull search},mendeley-groups={Cosijopii},pages={9},publisher={Association for Computing Machinery (ACM)},title={{cMOGA/D: a novel cellular GA based on decomposition to tackle constrained multiobjetive problems}},year={2020},month=jul}
2019
PMyS
Ca-PSO: Coulomb attracting Particle Swarms
Nayeli Joaquinita Meléndez Acosta, Ricardo Solano Monje, Cosijopii Garcia-Garcia, and
1 more author
Programación Matemática y Software, ISSN :2007-3283, Jul 2019
@article{GarciaGarcia2017,author={Garcia-Garcia, Cosijopii and {Mel{\'{e}}ndez Acosta}, Nayeli Joaquinita},booktitle={Advances in Social Informatics and its Applications},doi={10.13053/rcs-136-1-3},keywords={FCM Gender Classification Using Anthropometric Mea,LBP,Palabras clave: ANFIS,Texture Analysis and ANFIS Architecture,anfis,an{\'{a}}lisis de texturas,clasificaci{\'{o}}n,clasificaci{\'{o}}n de,clasificaci{\'{o}}n de g{\'{e}}nero,fcm,gender classification,g{\'{e}}nero,im{\'{a}}genes faciales,lbp,l{\'{o}}gica difusa,using anthropometric measurements},mendeley-groups={Bs Tesis,Cosijopii},pages={33--46},publisher={Research in Computing Science},title={{Clasificaci{\'{o}}n de g{\'{e}}nero utilizando medidas antropom{\'{e}}tricas, an{\'{a}}lisis de texturas y la arquitectura ANFIS}},volume={136},year={2017}}