Smooth Non-increasing Square Spatial Extents of Filters in Convolutional Layers of CNNs for Image Classification Problems
The present paper considers an open problem of setting hyperparameters for convolutional neural networks aimed at image classification. Since selecting filter spatial extents for convolutional layers is a topical problem, it is approximately solved by accumulating statistics of the neural network pe...
Main Author: | Romanuke Vadim V. |
---|---|
Format: | Article |
Language: | English |
Published: |
Sciendo
2018-05-01
|
Series: | Applied Computer Systems |
Subjects: | |
Online Access: | https://doi.org/10.2478/acss-2018-0007 |
Similar Items
-
An attempt of finding an appropriate number of convolutional layers in cnns based on benchmarks of heterogeneous datasets
by: Romanuke Vadim V.
Published: (2018-07-01) -
Appropriateness of Numbers of Receptive Fields in Convolutional Neural Networks Based on Classifying CIFAR-10 and EEACL26 Datasets
by: Romanuke Vadim
Published: (2018-12-01) -
EFFECT OF VARIOUS DIMENSION CONVOLUTIONAL LAYER FILTERS ON TRAFFIC SIGN CLASSIFICATION ACCURACY
by: V. N. Sichkar, et al.
Published: (2019-01-01) -
Improving the Separability of Deep Features with Discriminative Convolution Filters for RSI Classification
by: Na Liu, et al.
Published: (2018-03-01) -
Appropriate Number of Standard 2 × 2 Max Pooling Layers and Their Allocation in Convolutional Neural Networks for Diverse and Heterogeneous Datasets
by: Romanuke Vadim V.
Published: (2017-12-01)