1795. Neural Network Optimization using Evolutionary Computation
Invited abstract in session TB-34: Advancements of OR-analytics in statistics, machine learning and data science 3, stream Advancements of OR-analytics in statistics, machine learning and data science.
Tuesday, 10:30-12:00Room: Michael Sadler LG10
Authors (first author is the speaker)
| 1. | Ioannis Tsiligkaridis
|
| Heritage University |
Abstract
Optimizing a network model often becomes the primary interest, and if done manually, this can take some substantial effort. Evolutionary algorithms can provide with several optimization patterns that can solve a multitude of problems.
The Convolutional Neural Network (CNN) with the convolutional layer works as a filter that tends to excel at learning short sequences. A CNN architecture can be adapted using evolutionary algorithms. To this end GAs are used to encode a sequence of gene defining a CNN model for image classification.
Convolutional Autoencoders (CAE) is composed of the standard AE architecture and the convolutional layers. The CAE incorporates convolutional layers to better extract features in images. The construction process for individuals that works with encoder and decoder, includes the creation of the gene sequence and can check for additional layers and provide better performance for a multitude of problems.
Experiments with CAE providing the encoding of the architecture into a gene sequence shows the superiority for image classification over the CAE without using the Evolutionary Computation.
Keywords
- Analytics and Data Science
- Machine Learning
Status: accepted
Back to the list of papers