We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Strengthening the Training of Convolutional Neural Networks By Using Walsh Matrix

Abstract: DNN structures are continuously developing and achieving high performances in classification problems. Also, it is observed that success rates obtained with DNNs are higher than those obtained with traditional neural networks. In addition, one of the advantages of DNNs is that there is no need to spend an extra effort to determine the features; the CNN automatically extracts the features from the dataset during the training. Besides their benefits, the DNNs have the following three major drawbacks among the others: (i) Researchers have struggled with over-fitting and under-fitting issues in the training of DNNs, (ii) determination of even a coarse structure for the DNN may take days, and (iii) most of the time, the proposed network structure is too large to be too bulky to be used in real time applications. We have modified the training and structure of DNN to increase the classification performance, to decrease the number of nodes in the structure, and to be used with less number of hyper parameters. A minimum distance network (MDN) following the last layer of the convolutional neural network (CNN) is used as the classifier instead of a fully connected neural network (FCNN). In order to strengthen the training of the CNN, we suggest employing Walsh function. We tested the performances of the proposed DNN (named as DivFE) on the classification of ECG, EEG, heart sound, detection pneumonia in X-ray chest images, detection of BGA solder defects, and patterns of benchmark datasets (MNIST, IRIS, CIFAR10 and CIFAR20). In different areas, it has been observed that a higher classification performance was obtained by using the DivFE with less number of nodes.
Comments: Keyword: Deep neural networks, Convolutional neural network, Pattern recognition, Training deep neural network, Classification. arXiv admin note: substantial text overlap with arXiv:2103.10977
Subjects: Machine Learning (cs.LG)
Cite as: arXiv:2104.00035 [cs.LG]
  (or arXiv:2104.00035v1 [cs.LG] for this version)

Submission history

From: Tamer Olmez [view email]
[v1] Wed, 31 Mar 2021 18:06:11 GMT (2672kb)

Link back to: arXiv, form interface, contact.