Browsing by Subject "Artificial Neural Network"

Now showing 1 - 6 of 6

Open Access
Çarpmasız yapay sinir ağı
(IEEE, 2015-05) Akbaş, Cem Emre; Bozkurt, Alican; Çetin, A. Enis; Çetin-Atalay, R.; Üner, A.
Bu bildiride çarpma işlemi kullanmadan oluşturulan bir Yapay Sinir Ağı (YSA) sunulmaktadır. Girdi vektörleri ve YSA katsayılarının iç çarpımları çarpmasız bir vektör işlemiyle hesaplanmıştır. Yapay sinir ağının eğitimi sign-LMS algoritması ile yapılmıştır. Önerilen YSA sistemi, hesap gücü kısıtlı olan veya düşük enerji tüketimine ihtiyaç duyulan mikroişlemcilerde kullanılabilir.
Open Access
Comparison of the forecast performances of linear time series and artificial neural network models within the context of Turkish inflation
(2001) Uçar, Nuri
This thesis compares a variety of linear and nonlinear models to find the one with the best inflation forecast performance for the Turkish Economy. These comparisons are performed by considering the type of series whether or not stationary. Different combination techniques are applied to improve the forecasts. It is observed that the combination forecasts based on nonstationary vector autoregressive (VAR) and artificial neural network (ANN) models are better than the ones generated by other models. Furthermore, the forecast values combined with ANN technique produce lower root mean square errors (RMSE) than the other combination techniques.
Open Access
A fast neural-network algorithm for VLSI cell placement
(Pergamon Press, 1998) Aykanat, Cevdet; Bultan, T.; Haritaoğlu, İ.
Cell placement is an important phase of current VLSI circuit design styles such as standard cell, gate array, and Field Programmable Gate Array (FPGA). Although nondeterministic algorithms such as Simulated Annealing (SA) were successful in solving this problem, they are known to be slow. In this paper, a neural network algorithm is proposed that produces solutions as good as SA in substantially less time. This algorithm is based on Mean Field Annealing (MFA) technique, which was successfully applied to various combinatorial optimization problems. A MFA formulation for the cell placement problem is derived which can easily be applied to all VLSI design styles. To demonstrate that the proposed algorithm is applicable in practice, a detailed formulation for the FPGA design style is derived, and the layouts of several benchmark circuits are generated. The performance of the proposed cell placement algorithm is evaluated in comparison with commercial automated circuit design software Xilinx Automatic Place and Route (APR) which uses SA technique. Performance evaluation is conducted using ACM/SIGDA Design Automation benchmark circuits. Experimental results indicate that the proposed MFA algorithm produces comparable results with APR. However, MFA is almost 20 times faster than APR on the average.Cell placement is an important phase of current VLSI circuit design styles such as standard cell, gate array, and Field Programmable Gate Array (FPGA). Although nondeterministic algorithms such as Simulated Annealing (SA) were successful in solving this problem, they are known to be slow. In this paper, a neural network algorithm is proposed that produces solutions as good as SA in substantially less time. This algorithm is based on Mean Field Annealing (MFA) technique, which was successfully applied to various combinatorial optimization problems. A MFA formulation for the cell placement problem is derived which can easily be applied to all VLSI design styles. To demonstrate that the proposed algorithm is applicable in practice, a detailed formulation for the FPGA design style is derived, and the layouts of several benchmark circuits are generated. The performance of the proposed cell placement algorithm is evaluated in comparison with commercial automated circuit design software Xilinx Automatic Place and Route (APR) which uses SA technique. Performance evaluation is conducted using ACM/SIGDA Design Automation benchmark circuits. Experimental results indicate that the proposed MFA algorithm produces comparable results with APR. However, MFA is almost 20 times faster than APR on the average.
Open Access
Generalization and localization based style imitation for grayscale images
(Springer, Berlin, Heidelberg, 2003) Nar, F.; Çetin, Atılım
An example based rendering (EBR) method based on generalization and localization that uses artificial neural networks (ANN) and k-Nearest Neighbor (k-NN) is proposed. The method involves learning phase and application phase, which means that once a transformation filter is learned, it can be applied to any other image. In learning phase, error back-propagation learning algorithm is used to learn general transformation filter using unfiltered source image and filtered output image. ANNs are usually unable to learn filter-generated textures and brush strokes hence these localized features are stored in a feature instance table for using with k-NN during application phase. In application phase, for any given grayscale image, first ANN is applied then k-NN search is used to retrieve local features from feature instances considering texture continuity to produce desired image. Proposed method is applied up to 40 image filters that are collection of computer-generated and human-generated effects/styles. Good results are obtained when image is composed of localized texture/style features that are only dependent to intensity values of pixel itself and its neighbors.
Open Access
Innovation in the breeding of common bean through a combined approach of in vitro regeneration and machine learning algorithms
(Frontiers Media S.A., 2022-08-24) Aasim, Muhammad; Katirci, Ramazan; Baloch, Faheem Shehzad; Mustafa, Zemran; Bakhsh, Allahv; Nadeem, Muhammad Azhar; Ali, Seyid Amjad; Hatipoğlu, Rüştü; Çiftçi, Vahdettin; Habyarimana, Ephrem; Karaköy, Tolga; Chung, Yong Suk
Common bean is considered a recalcitrant crop for in vitro regeneration and needs a repeatable and efficient in vitro regeneration protocol for its improvement through biotechnological approaches. In this study, the establishment of efficient and reproducible in vitro regeneration followed by predicting and optimizing through machine learning (ML) models, such as artificial neural network algorithms, was performed. Mature embryos of common bean were pretreated with 5, 10, and 20 mg/L benzylaminopurine (BAP) for 20 days followed by isolation of plumular apice for in vitro regeneration and cultured on a post-treatment medium containing 0.25, 0.50, 1.0, and 1.50 mg/L BAP for 8 weeks. Plumular apice explants pretreated with 20 mg/L BAP exerted a negative impact and resulted in minimum shoot regeneration frequency and shoot count, but produced longer shoots. All output variables (shoot regeneration frequency, shoot counts, and shoot length) increased significantly with the enhancement of BAP concentration in the post-treatment medium. Interaction of the pretreatment × post-treatment medium revealed the need for a specific combination for inducing a high shoot regeneration frequency. Higher shoot count and shoot length were achieved from the interaction of 5 mg/L BAP × 1.00 mg/L BAP followed by 10 mg/L BAP × 1.50 mg/L BAP and 20 mg/L BAP × 1.50 mg/L BAP. The evaluation of data through ML models revealed that R2 values ranged from 0.32 to 0.58 (regeneration), 0.01 to 0.22 (shoot counts), and 0.18 to 0.48 (shoot length). On the other hand, the mean squared error values ranged from 0.0596 to 0.0965 for shoot regeneration, 0.0327 to 0.0412 for shoot count, and 0.0258 to 0.0404 for shoot length from all ML models. Among the utilized models, the multilayer perceptron model provided a better prediction and optimization for all output variables, compared to other models. The achieved results can be employed for the prediction and optimization of plant tissue culture protocols used for biotechnological approaches in a breeding program of common beans. Copyright © 2022 Aasim, Katirci, Baloch, Mustafa, Bakhsh, Nadeem, Ali, Hatipoğlu, Çiftçi, Habyarimana, Karaköy and Chung.
Open Access
Minyatür eylemsizlik duyucuları ve manyetometre sinyallerinin işlenmesiyle insan aktivitelerinin sınıflandırılması
(IEEE, 2011-04) Yüksek, Murat Cihan; Barshan, Billur
Bu çalışmada insan vücuduna yerleştirilen minyatür eylemsizlik duyucuları ve manyetometreler kullanılarak çeşitli aktiviteler örüntü tanıma yöntemleriyle ayırdedilmiş ve karşılaştırmalı bir çalışmanın sonuçları sunulmuştur. Ayırdetme işlemi için basit Bayeşçi (BB) yöntem, yapay sinir ağları (YSA), benzeşmezlik tabanlı sınıflandırıcı (BTS), ceşitli karar ağacı (KA) yöntemleri, Gauss karışım modeli (GKM) ve destek vektör makinaları (DVM) kullanılmıştır. Aktiviteler gövdeye, kollara ve bacaklara takılan beş duyucu ünitesinden gelen verilerin işlenmesiyle ayırdedilmiştir. Her ünite, her biri üç-eksenli olmak üzere birer ivmeölçer, dönüölçer ve manyetometre içermektedir. Çalışmanın sonuçlarına göre, en iyi ilk üç başarı oranı sırasıyla GKM (%99.12), YSA (%99.09) ve DVM (%98.90) yöntemleri ile elde edilmiştir.