scholarly journals Self-Improving Generative Artificial Neural Network for Pseudorehearsal Incremental Class Learning

Algorithms ◽  
2019 ◽  
Vol 12 (10) ◽  
pp. 206 ◽  
Author(s):  
Diego Mellado ◽  
Carolina Saavedra ◽  
Steren Chabert ◽  
Romina Torres ◽  
Rodrigo Salas
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Novelty Detection ◽  
Activation Function ◽  
Learning Sequence ◽  
Data Set ◽  
Detection Model ◽  
Neural Network System ◽  
Small Set ◽  
Artificial Neural

Deep learning models are part of the family of artificial neural networks and, as such, they suffer catastrophic interference when learning sequentially. In addition, the greater number of these models have a rigid architecture which prevents the incremental learning of new classes. To overcome these drawbacks, we propose the Self-Improving Generative Artificial Neural Network (SIGANN), an end-to-end deep neural network system which can ease the catastrophic forgetting problem when learning new classes. In this method, we introduce a novel detection model that automatically detects samples of new classes, and an adversarial autoencoder is used to produce samples of previous classes. This system consists of three main modules: a classifier module implemented using a Deep Convolutional Neural Network, a generator module based on an adversarial autoencoder, and a novelty-detection module implemented using an OpenMax activation function. Using the EMNIST data set, the model was trained incrementally, starting with a small set of classes. The results of the simulation show that SIGANN can retain previous knowledge while incorporating gradual forgetfulness of each learning sequence at a rate of about 7% per training step. Moreover, SIGANN can detect new classes that are hidden in the data with a median accuracy of 43 % and, therefore, proceed with incremental class learning.

scholarly journals Self-Improving Generative Artificial Neural Network for Pseudo-Rehearsal Incremental Class Learning

2019 ◽  
Author(s):  
Diego Mellado ◽  
Carolina Saavedra ◽  
Steren Chabert ◽  
Romina Torres ◽  
Rodrigo Salas
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Novelty Detection ◽  
Activation Function ◽  
Learning Sequence ◽  
Data Set ◽  
Detection Model ◽  
Neural Network System ◽  
Small Set ◽  
Artificial Neural

Deep learning models are part of the family of artificial neural networks and, as such, it suffers of catastrophic interference when they learn sequentially. In addition, most of these models have a rigid architecture which prevents the incremental learning of new classes. To overcome these drawbacks, in this article we propose the Self-Improving Generative Artificial Neural Network (SIGANN), a type of end-to-end Deep Neural Network system which is able to ease the catastrophic forgetting problem when leaning new classes. In this method, we introduce a novelty detection model to automatically detect samples of new classes, moreover an adversarial auto-encoder is used to produce samples of previous classes. This system consists of three main modules: a classifier module implemented using a Deep Convolutional Neural Network, a generator module based on an adversarial autoencoder; and a novelty detection module, implemented using an OpenMax activation function. Using the EMNIST data set, the model was trained incrementally, starting with a small set of classes. The results of the simulation show that SIGANN is able to retain previous knowledge with a gradual forgetfulness for each learning sequence. Moreover, SIGANN can detect new classes that are hidden in the data and, therefore, proceed with incremental class learning.

scholarly journals Run-off election-based decision method for the training and inference process in an artificial neural network

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Jingon Jang ◽  
Seonghoon Jang ◽  
Sanghyeon Choi ◽  
Gunuk Wang
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Decision Rule ◽  
Synaptic Weight ◽  
Activation Function ◽  
Network System ◽  
Run Off ◽  
Neural Network System ◽  
Artificial Neural ◽  
Fully Connected

AbstractGenerally, the decision rule for classifying unstructured data in an artificial neural network system depends on the sequence results of an activation function determined by vector–matrix multiplication between the input bias signal and the analog synaptic weight quantity of each node in a matrix array. Although a sequence-based decision rule can efficiently extract a common feature in a large data set in a short time, it can occasionally fail to classify similar species because it does not intrinsically consider other quantitative configurations of the activation function that affect the synaptic weight update. In this work, we implemented a simple run-off election-based decision rule via an additional filter evaluation to mitigate the confusion from proximity of output activation functions, enabling the improved training and inference performance of artificial neural network system. Using the filter evaluation selected via the difference among common features of classified images, the recognition accuracy achieved for three types of shoe image data sets reached ~ 82.03%, outperforming the maximum accuracy of ~ 79.23% obtained via the sequence-based decision rule in a fully connected single layer network. This training algorithm with an independent filter can precisely supply the output class in the decision step of the fully connected network.

scholarly journals USING APPLICATION OF AN ARTIFICIAL NEURAL NETWORK SYSTEM TO BACKCALCULATE PAVEMENT ELASTIC MODULUS

2020 ◽  
pp. 84-93
Author(s):  
M. M. M. Elshamy ◽  
A. N. Tiraturyan
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Elastic Moduli ◽  
Shock Loading ◽  
Back Propagation ◽  
Data Set ◽  
Neural Network System ◽  
Artificial Neural ◽  
Instrumental Recording ◽  
Accuracy Of Prediction

Statement of the problem. The article is devoted to the use of artificial neural networks in solving the problems of processing the results of instrumental recording of bowls of flexible pavement deflections using FWD shock loading settings. Results. The analysis was carried out, the shortcomings of the existing processing methods were noted, in particular the “backcalculation” method, which consists of a long calculation time, and the instability of the results obtained. The structure of the artificial neural network was built to determine the elastic moduli of the pavement layers. Training of an artificial neural network was carried out using the method of back propagation of error. Conclusions. The developed neural network has shown good results in training on the test data set, as well as high accuracy of prediction of the elastic moduli of the pavement.

scholarly journals Application of Artificial Neural Network to Predict Biodiesel Yield from Waste Frying Oil Transesterification

2020 ◽  
Vol 5 (1) ◽  
pp. 62-74
Author(s):  
Agus Haryanto ◽  
Tri Wahyu Saputra ◽  
Mareli Telaumbanua ◽  
Amiera Citra Gita
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Reaction Time ◽  
Activation Function ◽  
Frying Oil ◽  
Biodiesel Production ◽  
Ann Model ◽  
Data Set ◽  
Ann Models ◽  
Artificial Neural

Used frying oil (UFO) has a great potential as feedstock for biodiesel production. This study aims to develop an artificial neural  network  (ANN)  model  to  predict  biodiesel  yield produced from base-catalyzed transesterification of UFO. The experiment  was  performed  with  100  mL  of  UFO  at  three different  molar  ratios  (oil:methanol) (namely 1:4,  1:5,  and 1:6), conducted with reaction temperatures of  30 to 55oC (raised by 5oC), and reaction time of 0.25, 0.5, 1, 2, 3, 6, 8, and 10 minutes. Prediction model was based on ANN model consisting  of  three  layers  with  27  combinations  of  three activation  functions  (tansig,  logsig,  purelin).  All  activation function  architectures  were  trained  using  Levenberg- Marquardt train type with 126 data set (87.5%) and learning rate  of  0.001.  Model  validation  used  18  data  set  (12.5%) measured at reaction time of 8 min. Results showed that two ANN models with activation function of logsig-purelin-logsig and purelin-logsig-tansig be the best with RRMSE of 2.41% and  2.44%  with  R2  of  0.9355  and  0.9391,  respectively. Predictions   of   biodiesel   yield   using   ANN   models   are significantly better than those of first-order kinetics.

Random Forest (RF) and Artificial Neural Network (ANN) Algorithms for LULC Mapping

2020 ◽  
Vol 38 (4A) ◽  
pp. 510-514
Author(s):  
Tay H. Shihab ◽  
Amjed N. Al-Hameedawi ◽  
Ammar M. Hamza
Keyword(s):  
Neural Network ◽  
Remote Sensing ◽  
Artificial Neural Network ◽  
Random Forest ◽  
Satellite Image ◽  
Landsat 8 ◽  
Optical Remote Sensing ◽  
Data Set ◽  
Artificial Neural ◽  
Artificial Neural Network Ann

In this paper to make use of complementary potential in the mapping of LULC spatial data is acquired from LandSat 8 OLI sensor images are taken in 2019.  They have been rectified, enhanced and then classified according to Random forest (RF) and artificial neural network (ANN) methods. Optical remote sensing images have been used to get information on the status of LULC classification, and extraction details. The classification of both satellite image types is used to extract features and to analyse LULC of the study area. The results of the classification showed that the artificial neural network method outperforms the random forest method. The required image processing has been made for Optical Remote Sensing Data to be used in LULC mapping, include the geometric correction, Image Enhancements, The overall accuracy when using the ANN methods 0.91 and the kappa accuracy was found 0.89 for the training data set. While the overall accuracy and the kappa accuracy of the test dataset were found 0.89 and 0.87 respectively.

scholarly journals Prediction of Stress in Power Transformer Winding Conductors Using Artificial Neural Networks: Hyperparameter Analysis

Energies ◽  
2021 ◽  
Vol 14 (14) ◽  
pp. 4242
Author(s):  
Fausto Valencia ◽  
Hugo Arcos ◽  
Franklin Quilumba
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Finite Element ◽  
Power Transformer ◽  
Network Models ◽  
Activation Function ◽  
Neural Network Models ◽  
Transformer Winding ◽  
Artificial Neural ◽  
Femm Software

The purpose of this research is the evaluation of artificial neural network models in the prediction of stresses in a 400 MVA power transformer winding conductor caused by the circulation of fault currents. The models were compared considering the training, validation, and test data errors’ behavior. Different combinations of hyperparameters were analyzed based on the variation of architectures, optimizers, and activation functions. The data for the process was created from finite element simulations performed in the FEMM software. The design of the Artificial Neural Network was performed using the Keras framework. As a result, a model with one hidden layer was the best suited architecture for the problem at hand, with the optimizer Adam and the activation function ReLU. The final Artificial Neural Network model predictions were compared with the Finite Element Method results, showing good agreement but with a much shorter solution time.

Watermelon Sorting Process by Frequency Identification and Artificial Neural Network

2021 ◽  
Author(s):  
Komsan Wongkalasin ◽  
Teerapon Upachaban ◽  
Wacharawish Daosawang ◽  
Nattadon Pannucharoenwong ◽  
Phadungsak Ratanadecho
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Selection Process ◽  
Network Processor ◽  
Data Set ◽  
The Neural Network ◽  
Frequency Identification ◽  
Sorting Process ◽  
Artificial Neural ◽  
Reasonable Prediction

This research aims to enhance the watermelon’s quality selection process, which was traditionally conducted by knocking the watermelon fruit and sort out by the sound’s character. The proposed method in this research is generating the sound spectrum through the watermelon and then analyzes the response signal’s frequency and the amplitude by Fast Fourier Transform (FFT). Then the obtained data were used to train and verify the neural network processor. The result shows that, the frequencies of 129 and 172 Hz were suit to be used in the comparison. Thirty watermelons, which were randomly selected from the orchard, were used to create a data set, and then were cut to manually check and match to the fruits’ quality. The 129 Hz frequency gave the response ranging from 13.57 and above in 3 groups of watermelons quality, including, not fully ripened, fully ripened, and close to rotten watermelons. When the 172 Hz gave the response between 11.11–12.72 in not fully ripened watermelons and those of 13.00 or more in the group of close to rotten and hollow watermelons. The response was then used as a training condition for the artificial neural network processor of the sorting machine prototype. The verification results provided a reasonable prediction of the ripeness level of watermelon and can be used as a pilot prototype to improve the efficiency of the tools to obtain a modern-watermelon quality selection tool, which could enhance the competitiveness of the local farmers on the product quality control.

A Model on the Correlation between Composition and Mechanical Properties of Mg-Al-Zn Alloys by Using Artificial Neural Network

2005 ◽  
Vol 488-489 ◽  
pp. 793-796 ◽  
Author(s):  
Hai Ding Liu ◽  
Ai Tao Tang ◽  
Fu Sheng Pan ◽  
Ru Lin Zuo ◽  
Ling Yun Wang
Keyword(s):  
Neural Network ◽  
Mechanical Properties ◽  
Artificial Neural Network ◽  
Magnesium Alloys ◽  
Tensile Yield Strength ◽  
Data Set ◽  
The Neural Network ◽  
Artificial Neural ◽  
Prediction Of Mechanical Properties ◽  
Artificial Neural Network Ann

A model was developed for the analysis and prediction of correlation between composition and mechanical properties of Mg-Al-Zn (AZ) magnesium alloys by applying artificial neural network (ANN). The input parameters of the neural network (NN) are alloy composition. The outputs of the NN model are important mechanical properties, including ultimate tensile strength, tensile yield strength and elongation. The model is based on multilayer feedforward neural network. The NN was trained with comprehensive data set collected from domestic and foreign literature. A very good performance of the neural network was achieved. The model can be used for the simulation and prediction of mechanical properties of AZ system magnesium alloys as functions of composition.

Sign in / Sign up

Export Citation Format

Share Document