NON-EUCLIDEAN VECTOR PRODUCT FOR NEURAL NETWORKS

Date

2018-04-20

Author

Afrasiyabi, Arman
Badawi, Diaa
Nasır, Barış
Yildiz, Ozan
Yarman Vural, Fatoş Tunay
ÇETİN, AHMET ENİS

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

132
views

0
downloads

We present a non-Euclidean vector product for artificial neural networks. The vector product operator does not require any multiplications while providing correlation information between two vectors. Ordinary neurons require inner product of two vectors. We propose a class of neural networks with the universal approximation property over the space of Lebesgue integrable functions based on the proposed non-Euclidean vector product. In this new network, the "product" of two real numbers is defined as the sum of their absolute values, with the sign determined by the sign of the product of the numbers. This "product' is used to construct a vector product in R-N. The vector product induces the l(1) norm. The additive neural network successfully solves the XOR problem. Experiments on MNIST and CIFAR datasets show that the classification performance of the proposed additive neural network is comparable to the corresponding multi-layer perceptron and convolutional neural networks.

Subject Keywords

Non-Euclidean operator, Additive neural networks, Multiplication-free operator

URI

https://hdl.handle.net/11511/55281

Conference Name

IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Collections

Department of Computer Engineering, Conference / Seminar

Suggestions

OpenMETU
Core

Asymptotic Behavior of Lotz-Rabiger and Martingale Nets Emelyanov, Eduard (2010-09-01) Using Theorem 1 (of convergence) in [1], we prove several results on LR- and M-nets by a unified approach to these nets that appear as the two extreme types of asymptotically abelian nets.
Representing temporal knowledge in connectionist expert systems Alpaslan, Ferda Nur (1996-09-27) This paper introduces a new temporal neural networks model which can be used in connectionist expert systems. Also, a Variation of backpropagation algorithm, called the temporal feedforward backpropagation algorithm is introduced as a method for training the neural network. The algorithm was tested using training examples extracted from a medical expert system. A series of experiments were carried out using the temporal model and the temporal backpropagation algorithm. The experiments indicated that the alg...
Domain-Structured Chaos in a Hopfield Neural Network Akhmet, Marat (World Scientific Pub Co Pte Lt, 2019-12-30) In this paper, we provide a new method for constructing chaotic Hopfield neural networks. Our approach is based on structuring the domain to form a special set through the discrete evolution of the network state variables. In the chaotic regime, the formed set is invariant under the system governing the dynamics of the neural network. The approach can be viewed as an extension of the unimodality technique for one-dimensional map, thereby generating chaos from higher-dimensional systems. We show that the dis...
Dual-Band Antenna Array Optimizations Using Heuristic Algorithms and the Multilevel Fast Multipole Algorithm Onol, Can; Gokce, Ozer; Ergül, Özgür Salih (2015-05-24) We consider design and simulations of dual-band antenna arrays and their optimizations via heuristic algorithms, particularly, genetic algorithms (GAs) and particle swarm optimization (PSO) methods. As shown below, these arrays consist of patch antennas of different sizes, depending on the target frequencies. The resulting radiation problems are solved iteratively, where the matrix-vector multiplications are performed efficiently with the multilevel fast multipole algorithm (MLFMA). MLFMA allows for realist...
Binarized Weight Networks for Inverse Problems Ozkan, Savas; Becek, Kadircan; Inci, Alperen; Kutukcu, Basar; Ugurcali, Faruk; Kaya, Mete Can; Akar, Gözde (2020-01-01) In this paper, we present a binarized neural network structure for inverse problems. In this structure, memory requirements and computation time are significantly reduced with a negligible performance drop compared to full-precision models. For this purpose, a unique architecture is proposed based on a residual learning. Precisely, it opts to reconstruct only the error between input and output images, which is eventually centralized the responses around zero. To this end, this provides several advantages fo...

Citation Formats

A. Afrasiyabi, D. Badawi, B. Nasır, O. Yildiz, F. T. Yarman Vural, and A. E. ÇETİN, “NON-EUCLIDEAN VECTOR PRODUCT FOR NEURAL NETWORKS,” presented at the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Calgary, CANADA, 2018, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/55281.