-
ResVGAE: Going Deeper with Residual Modules for Link Prediction
Authors:
Indrit Nallbani,
Reyhan Kevser Keser,
Aydin Ayanzadeh,
Nurullah Çalık,
Behçet Uğur Töreyin
Abstract:
Graph autoencoders are efficient at embedding graph-based data sets. Most graph autoencoder architectures have shallow depths which limits their ability to capture meaningful relations between nodes separated by multi-hops. In this paper, we propose Residual Variational Graph Autoencoder, ResVGAE, a deep variational graph autoencoder model with multiple residual modules. We show that our multiple…
▽ More
Graph autoencoders are efficient at embedding graph-based data sets. Most graph autoencoder architectures have shallow depths which limits their ability to capture meaningful relations between nodes separated by multi-hops. In this paper, we propose Residual Variational Graph Autoencoder, ResVGAE, a deep variational graph autoencoder model with multiple residual modules. We show that our multiple residual modules, a convolutional layer with residual connection, improve the average precision of the graph autoencoders. Experimental results suggest that our proposed model with residual modules outperforms the models without residual modules and achieves similar results when compared with other state-of-the-art methods.
△ Less
Submitted 4 August, 2022; v1 submitted 3 May, 2021;
originally announced May 2021.
-
PURSUhInT: In Search of Informative Hint Points Based on Layer Clustering for Knowledge Distillation
Authors:
Reyhan Kevser Keser,
Aydin Ayanzadeh,
Omid Abdollahi Aghdam,
Caglar Kilcioglu,
Behcet Ugur Toreyin,
Nazim Kemal Ure
Abstract:
One of the most efficient methods for model compression is hint distillation, where the student model is injected with information (hints) from several different layers of the teacher model. Although the selection of hint points can drastically alter the compression performance, conventional distillation approaches overlook this fact and use the same hint points as in the early studies. Therefore,…
▽ More
One of the most efficient methods for model compression is hint distillation, where the student model is injected with information (hints) from several different layers of the teacher model. Although the selection of hint points can drastically alter the compression performance, conventional distillation approaches overlook this fact and use the same hint points as in the early studies. Therefore, we propose a clustering based hint selection methodology, where the layers of teacher model are clustered with respect to several metrics and the cluster centers are used as the hint points. Our method is applicable for any student network, once it is applied on a chosen teacher network. The proposed approach is validated in CIFAR-100 and ImageNet datasets, using various teacher-student pairs and numerous hint distillation methods. Our results show that hint points selected by our algorithm results in superior compression performance compared to state-of-the-art knowledge distillation algorithms on the same student models and datasets.
△ Less
Submitted 3 November, 2022; v1 submitted 26 February, 2021;
originally announced March 2021.
-
Gaussian Three-Dimensional kernel SVM for Edge Detection Applications
Authors:
Safar Irandoust-Pakchin,
Aydin Ayanzadeh,
Siamak Beikzadeh
Abstract:
This paper presents a novel and uniform algorithm for edge detection based on SVM (support vector machine) with Three-dimensional Gaussian radial basis function with kernel. Because of disadvantages in traditional edge detection such as inaccurate edge location, rough edge and careless on detect soft edge. The experimental results indicate how the SVM can detect edge in efficient way. The performa…
▽ More
This paper presents a novel and uniform algorithm for edge detection based on SVM (support vector machine) with Three-dimensional Gaussian radial basis function with kernel. Because of disadvantages in traditional edge detection such as inaccurate edge location, rough edge and careless on detect soft edge. The experimental results indicate how the SVM can detect edge in efficient way. The performance of the proposed algorithm is compared with existing methods, including Sobel and canny detectors. The results show that this method is better than classical algorithm such as canny and Sobel detector.
△ Less
Submitted 30 September, 2017;
originally announced October 2017.
