Accès libre

Designing Convolutional Neural Network Architecture Using Genetic Algorithms

Ashray Bhandare

Ashray Bhandare

Department of EECS, the University of ToledoToledo, USA
Recherchez cet auteur sur
Sciendo|Google Scholar
Bhandare, Ashray
 et 
Devinder Kaur

Devinder Kaur

Department of EECS, the University of ToledoToledo, USA
Recherchez cet auteur sur
Sciendo|Google Scholar
Kaur, Devinder
  
22 févr. 2021
Designing Convolutional Neural Network Architecture Using Genetic Algorithms's Cover Image
À propos de cet article
Article précédent
Article suivant
Citez
Télécharger la couverture
Publié en ligne: 22 févr. 2021
Pages: 26 - 35
DOI: https://doi.org/10.21307/ijanmc-2021-024
Mots clés
© 2021 Ashray Bhandare et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1.

An example of CNN architecture [10]
An example of CNN architecture [10]
undefined

Figure 2.

The process of convolution on the left-hand corner of an image
The process of convolution on the left-hand corner of an image

Figure 3.

Population of three chromosomes
Population of three chromosomes

Figure 4.

Single-point crossover
Single-point crossover

Figure 5.

Two-point crossover
Two-point crossover

Figure 6.

Representation of a GA chromosome
Representation of a GA chromosome

Figure 7.

Representation of the hyper-parameters in binary format
Representation of the hyper-parameters in binary format

Figure 8.

Generated CNN architecture after GA tuning
Generated CNN architecture after GA tuning

Parameters of the genetic operations

Parameters Value
Tournament selection size 2
Crossover Probability 50%
Mutation probability 80%
Genes Mutated 10%

Highest fitness values obtained during each of the 10 experiments

Exp. No. Highest Fitness Value
1 0.984499992943
2 0.973899998105
3 0.988800008184
4 0.991900001359
5 0.947799991965
6 0.949000005102
7 0.983099997652
8 0.979799999475
9 0.956399999567
10 0.972350000068

The various hyper parameters in CNN with their ranges

Hyper parameter Range
No. of Epoch (0 – 127)
Batch Size (0 – 256)
No. of Convolution Layers (0 – 8)
No. of Filters at each Convo layer (0 – 64)
Convo Filter Size at each Convo layer (0 – 8)
Activations used at each Convo layer (sigmoid, tanh, relu, linear)
Maxpool layer after each Convo layer (true, false)
Maxpool Pool Size for each Maxpool layer (0 – 8)
No. of Feed-Forward Hidden Layers (0 – 8)
No. of Feed-Forward Hidden Neurons at each layer (0 – 64)
Activations used at each Feed-Forward layer (sigmoid, tanh, softmax, relu)
Optimizer (Adagrad, Adadelta, RMS, SGD)
Langue:
Anglais
Périodicité:
4 fois par an
Sujets de la revue:
Informatique, Informatique, autres
RSS Feed de la revue