Adaptive Neural Network Model Research Articles

BiPhase adaptive learning-based neural network model for cloud datacenter workload forecasting

Cloud computing promises elasticity, flexibility and cost-effectiveness to satisfy service level agreement conditions. The cloud service providers should plan and provision the computing resources rapidly to ensure the availability of infrastructure to match the demands with closed proximity. The workload prediction has become critical as it can be helpful in managing the infrastructure effectively. In this paper, we present a workload forecasting framework based on neural network model with supervised learning technique. An improved and adaptive differential evolution algorithm is developed to improve the learning efficiency of predictive model. The algorithm is capable of optimizing the best suitable mutation operator and crossover operator. The prediction accuracy and convergence rate of the learning are observed to be improved due to its adaptive behavior in pattern learning from sampled data. The predictive model’s performance is evaluated on four real-world data traces including Google cluster trace and NASA Kennedy Space Center logs. The results are compared with state-of-the-art methods, and improvements up to 91%, 97% and 97.2% are observed over self-adaptive differential evolution, backpropagation and average-based workload prediction techniques, respectively.

A Simple 3-Parameter Model for Examining Adaptation in Speech and Voice Production.

Sensorimotor adaptation experiments are commonly used to examine motor learning behavior and to uncover information about the underlying control mechanisms of many motor behaviors, including speech production. In the speech and voice domains, aspects of the acoustic signal are shifted/perturbed over time via auditory feedback manipulations. In response, speakers alter their production in the opposite direction of the shift so that their perceived production is closer to what they intended. This process relies on a combination of feedback and feedforward control mechanisms that are difficult to disentangle. The current study describes and tests a simple 3-parameter mathematical model that quantifies the relative contribution of feedback and feedforward control mechanisms to sensorimotor adaptation. The model is a simplified version of the DIVA model, an adaptive neural network model of speech motor control. The three fitting parameters of SimpleDIVA are associated with the three key subsystems involved in speech motor control, namely auditory feedback control, somatosensory feedback control, and feedforward control. The model is tested through computer simulations that identify optimal model fits to six existing sensorimotor adaptation datasets. We show its utility in (1) interpreting the results of adaptation experiments involving the first and second formant frequencies as well as fundamental frequency; (2) assessing the effects of masking noise in adaptation paradigms; (3) fitting more than one perturbation dimension simultaneously; (4) examining sensorimotor adaptation at different timepoints in the production signal; and (5) quantitatively predicting responses in one experiment using parameters derived from another experiment. The model simulations produce excellent fits to real data across different types of perturbations and experimental paradigms (mean correlation between data and model fits across all six studies = 0.95 ± 0.02). The model parameters provide a mechanistic explanation for the behavioral responses to the adaptation paradigm that are not readily available from the behavioral responses alone. Overall, SimpleDIVA offers new insights into speech and voice motor control and has the potential to inform future directions of speech rehabilitation research in disordered populations. Simulation software, including an easy-to-use graphical user interface, is publicly available to facilitate the use of the model in future studies.

Comparison of Power Output Forecasting on the Photovoltaic System Using Adaptive Neuro-Fuzzy Inference Systems and Particle Swarm Optimization-Artificial Neural Network Model

The power output forecasting of the photovoltaic (PV) system is essential before deciding to install a photovoltaic system in Nakhon Ratchasima, Thailand, due to the uneven power production and unstable data. This research simulates the power output forecasting of PV systems by using adaptive neuro-fuzzy inference systems (ANFIS), comparing accuracy with particle swarm optimization combined with artificial neural network methods (PSO-ANN). The simulation results show that the forecasting with the ANFIS method is more accurate than the PSO-ANN method. The performance of the ANFIS and PSO-ANN models were verified with mean square error (MSE), root mean square error (RMSE), mean absolute error (MAP) and mean absolute percent error (MAPE). The accuracy of the ANFIS model is 99.8532%, and the PSO-ANN method is 98.9157%. The power output forecast results of the model were evaluated and show that the proposed ANFIS forecasting method is more beneficial compared to the existing method for the computation of power output and investment decision making. Therefore, the analysis of the production of power output from PV systems is essential to be used for the most benefit and analysis of the investment cost.

Analysis of the Smart Player’s Impact on the Success of a Team Empowered with Machine Learning

The innovation and development in data science have an impact in all trades of life. The commercialization of sport has encouraged players, coaches, and other concerns to use technology to be in better position than r their opponents. In the past, the focus was on improved training techniques for better physical performance. These days, sports analytics identify the patterns in the performance and highlight strengths and weaknesses of potential players. Sports analytics not only predict the performance of players in the near future but it also performs predictive modeling for a particular behavior of a player in the past. The impact of a smart player on the success of a team is always a big question mark before the start of a match. The fans always want to know performance analysis of these superstar players and they always are interested to get to know more about their favorite player and they always have high hopes from their favorite player. Machine learning (ML) based techniques help in predicting the performance of an individual player as well as for the whole team. The statistics are very vital and useful for management, fans, and expert analysis. In our proposed framework, the adaptive back propagation neural network (ABPNN) model is used for the prediction of a player’s performance. The data is collected from football websites, and the results are stored in the cloud for fast fetching of data. They can be retrieved anywhere in the world through cloud storage. The results are computed with 94% accuracy and the performance of the smart player is formulated for the success of a team.

An Adaptive Vector Median Filtering Approach to Enhance the Prediction Efficiency of Signal Power Loss

This research work designed and implemented an adaptive Artificial Neural Network (ANN) model using Multi-Layer Perceptron (MLP) and Radial Basis Function (RBF) models built on a Vector Median Filter (VMF) for pre-processing of the dataset. Normalized dataset is denoised using VMF and trained with both MLP- and RBF-ANN models. The proposed model has been developed from measurement data collected from two transmitter locations of non-line-of-sight and line-of-sight operating at the 1900MHz frequency band from LTE cellular network over distances of 1800m 1400m respectively. For non-line-of-sight site-1, VMF-MLP gives a correlation coefficient of 0.9600 compared to 0.9490 for VMF-RBF with a Bayesian regularization training algorithm. The VMF-MLP has 2.1380, 1.5000, and 1.4510 for root mean squared error, mean absolute error, and standard deviation compared to 2.3550, 1.5370, and 1.5610 for VMF-RBF network, respectively. The same trend was seen for line-of-sight in site-2 where correlation coefficient for VMF-MLP is 0.9900 and for VMF-RBF is 0.9840. The VMF-MLP has root mean squared error, mean absolute error, and standard deviation as 2.0670, 1.4900, and 1.3180, respectively, compared to VMF-RBF as 2.3470, 1.9010, and 1.3760, respectively. The predictions of these measurement data have been analyzed in this research work.

A survey of adaptive resonance theory neural network models for engineering applications

This survey samples from the ever-growing family of adaptive resonance theory (ART) neural network models used to perform the three primary machine learning modalities, namely, unsupervised, supervised and reinforcement learning. It comprises a representative list from classic to contemporary ART models, thereby painting a general picture of the architectures developed by researchers over the past 30 years. The learning dynamics of these ART models are briefly described, and their distinctive characteristics such as code representation, long-term memory, and corresponding geometric interpretation are discussed. Useful engineering properties of ART (speed, configurability, explainability, parallelization and hardware implementation) are examined along with current challenges. Finally, a compilation of online software libraries is provided. It is expected that this overview will be helpful to new and seasoned ART researchers.

Adaptive indirect neural network model for roughness in honing processes

Abstract Honing processes provide a crosshatch pattern that allows oil flow, for example in combustion engine cylinders. This paper provides an adaptive neural network model for predicting roughness as a function of process parameters. Input variables are three parameters from the Abbott-Firestone curve, Rk, Rpk and Rvk. Output parameters are grain size, density of abrasive, pressure, linear speed and tangential speed. The model consists of applying a direct and an indirect model consecutively, with one convergence parameter and one error parameter. The indirect model has one network with 48 neurons and the direct model has three networks having 25, 9 and 5 neurons respectively. The adaptive one allows selecting discrete values for some variables like grain size or density.

Adaptive convolutional neural network using N-gram for spatial object recognition

Remote sensing applications are playing a vital role to improve the commercial satellite imagery with high resolution. In the spatial information system, object detection is the basic needs for computing the mathematical model. Geographical object related analysis for the image is used to gather data from remote sensing images. In this paper, we propose an Adaptive Convolutional Neural Network model using N-gram for Spatial Object Recognition on Satellite Images. Our methodology needs a learning model for the structures in the images to gather the data using prior knowledge. N-gram uses the functionalities of learning models. Spatial object recognition is performed using the learning method to segment the images with the human subjects that can increase their understanding of including the perception, cognition and decision. The result obtained for two stage of image processing is collected, and a relationship to psychological and mathematical basis is made. The results show that convinced association relevant level to the human perception is serving additional to identify the spatial objects. The experimentation is performed in MATLAB software where the results proved that our methodology is superior suitable for precise object detection and recognition on dissimilar levels of satellite images.

Comparison of neural network and neurofuzzy identification of vehicle handling under uncertainties

In this study, two adaptive neural network and neurofuzzy identification models are proposed to identify vehicle handling under uncertainties. These models are used to identify vehicle handling in different road friction coefficients and velocities. These two identification models modify their weights to cope with uncertainties using back propagation of error as a learning algorithm. However, an adaptive model has some limitations to identify real systems. The ability of adaptation is not the same for all identification models; some models are more robust to cope with a specific uncertainty or a wider range of uncertainties. In this study, adaptiveness of two identification models are compared under two different uncertainties. First, a precise model in CARSIM software is simulated and a set of input/output data of vehicle response are collected. Then an initial three-layer neural network is trained in MATLAB software. In addition, a Neurofuzzy model is also trained in ANFIS (adaptive neurofuzzy inference system) toolbox of MATLAB software. Then this trained model is applied to the vehicle in different maneuvers, velocities and road friction coefficients. Results show that proposed neural network identifies the vehicle handling more efficiently than neurofuzzy model in conditions that are away from training condition. However, proposed neurofuzzy model is more precise and accurate than neural network in the training condition.

Stress wave evaluation for predicting the properties of thermally modified wood using neuro-fuzzy and neural network modeling

AbstractThis study investigated using the stress wave method to predict the properties of thermally modified wood by means of an adaptive neuro-fuzzy inference system (ANFIS) and neural network (NN) modeling. The stress wave was detected using a pair of accelerometers and an acoustic emission (AE) sensor, and the effect of heat treatment (HT) on the physical and mechanical properties of wood as well as wave velocity and AE signal is discussed. The AE signal was processed in the time and time-frequency domains using wavelet analysis and different features were extracted for network training. The auto-associative NN is used as a dimensional reduction method to decrease the dimension of the extracted AE features and enhance the ANFIS performance. It was shown that while the stress wave velocity using the accelerometer did not result in an accurate model, the network performance significantly increased when trained with the AE features. The AE signal exhibited a significant correlation with wood treatment and porosity. The best ANFIS performance corresponded to predicting the wood swelling coefficient, equilibrium moisture content (EMC) and water absorption (WA), respectively. However, the AE signal did not seem suitable for predicting the wood density and hardness. The performance of ANFIS was compared with the “group method of data handling” (GMDH) NN. Both the ANFIS and GMDH networks showed higher accuracy than the multivariate linear regression (MVLR) model.

Below the Data Range Prediction of Soft Computing Wave Reflection of Semicircular Breakwater

Coastal defenses such as the breakwaters are important structures to maintain the navigation conditions in a harbor. The estimation of their hydrodynamic characteristics is conventionally done using physical models, subjecting to higher costs and prolonged procedures. Soft computing methods prove to be useful tools, in cases where the data availability from physical models is limited. The present paper employs adaptive neuro-fuzzy inference system (ANFIS) and artificial neural network (ANN) models to the data obtained from physical model studies to develop a novel methodology to predict the reflection coefficient (Kr) of seaside perforated semicircular breakwaters under low wave heights, for which no physical model data is available. The prediction was done using the input parameters viz., incident wave height (Hi), wave period (T), center-to-center spacing of perforations (S), diameter of perforations (D), radius of semicircular caisson (R), water depth (d), and semicircular breakwater structure height (hs). The study shows the prediction below the available data range of wave heights is possible by ANFIS and ANN models. However, the ANFIS performed better with R2 = 0.9775 and the error reduced in comparison with the ANN model with R2 = 0.9751. Study includes conventional data segregation and prediction using ANN and ANFIS.

Probabilistic Adaptive Genetic Algorithm for Pre-treatment Radiomics Features Predict Patient Survival

In order to provide personalized treatment for patients with tongue carcinoma, a probabilistic adaptive genetic algorithm neural network (PAGA-BP) model is proposed in this paper. The PAGA-BP model ameliorates selection sorting operator, adaptive crossover operator and u-adaptive mutation operator to optimize the initial weight of BP neural network. By comparing with traditional GA-BP neural network and BP neural network survival prediction model, the results show that PAGA-BP prediction model has the highest approximation accuracy and better survival period prediction.

Brain Image Recognition Algorithm and High Performance Computing of Internet of Medical Things Based on Convolutional Neural Network

Due to the wide variety of medical images and the complexity of the human body structure, the characteristics of manual extraction of medical images are difficult, the adaptive ability is poor, and the classification effect needs to be improved. Aiming at the shortcomings of traditional medical image recognition methods, this paper proposes an adaptive convolutional neural network model CNN-BN-PReLU based on the convolutional neural network method. The model first performs batch normalization (BN) processing on the input of each feature map of each layer of network, and then adaptively adjusts the parameters by using Parametric Rectified Linear Unit (PReLU) to compare the BN algorithm. Based on the performance before and after the activation function, an adaptive convolutional neural network model is constructed. The experimental results show that the model can abstract the image features without artificial intervention, speed up the network convergence speed and shorten the training time, and significantly improve the image recognition rate and reduce the misdiagnosis rate and missed diagnosis rate of the disease.

Modeling, Monitoring and Prediction of Drought in Iran

The drought phenomenon is not specific to the region and it affects different parts of the world. One of these areas is Iran in Southwest Asia, which suffered from this phenomenon in recent years. The purpose of this study is to model, analyze and predict the drought in Iran. To do this, climatic parameters (precipitation, temperature, sunshine, minimum relative humidity and wind speed) were used at 30 stations in the period of 29 years (1990-2018). For modeling of TIBI fuzzy index, first, four indicators (SET, SPI, SEB, MCZI) were fuzzy in Matlab software. Then the indices were compared and Topsis model were used for prioritizing areas involved with drought. Finally, Anfis adaptive artificial neural network model was used to predict. Results showed that the new fuzzy index TIBI for classifying drought reflected four of the above indicators with high accuracy. Among these five climatic parameters used in this study, the temperature and precipitation parameters had the most influential effect on the fluctuation of drought severity. The severity of drought was more based on a 6-month scale modeling than 12 months. The highest percentage of drought occurrence was at Bandar Abbas station with a value of 24.3 on a 12-month scale and the lowest was in Shahrekord station with a percentage of 0.36% on a six-month scale. Based on Anfis model and TIBI fuzzy index, Bandar Abbas, Bushehr and Zahedan stations were more exposed to drought due to the TIBI index of 0.62, 0.96 and 0.97, respectively. According to the results in both 6 and 12 months scale, the southern regions of Iran were more severely affected by drought, which requires suitable water management in these areas.

Neural Network Based Luminance Variation Resistant Remote-Photoplethysmography for Driver’s Heart Rate Monitoring

The driver's physiological status has enormous value to public traffic safety and cannot be ignored nowadays. In addition, heart rate (HR) is one of the most important indicators of human's health status. When detecting the driver's HR, using traditional contact-type devices might bring about the driver's distraction or discomfort. On the contrary, the remote photoplethysmography (rPPG) technique is a better way to monitor a driver's HR in vehicle applications simply by using a web-camera without interfering the driver. Most of the rPPG studies intended to reduce the interference caused by facial motion or luminance changes in the indoor or controlled scenario, but there are relatively fewer discussions on outdoor scenarios. Consequently, the purpose of this paper is to enhance the rPPG technique to make it suitable for the outdoor driving scenarios and for monitoring the driver's HR in different weather conditions, including daytime and nighttime. We first utilize artificial neural network (ANN) and train multiple personalized ANN models for each driver. For predicting the drivers' HR beat more precisely, we propose the approach, adaptive neural network model selection (ANNMS), which adaptively selects a personalized ANN model based on different noise conditions. Our algorithm eliminates the effect of noises caused by the variations of facial luminance in eight outdoor driving scenarios. The proposed driver's HR beat monitoring system has been evaluated against the state-of-the-art rPPG techniques that are Chrominance signal-based (CHRO) and k-nearest neighbors-based (kNN) algorithms. Compared with the CHRO and kNN algorithms, the ANNMS reduces the mean absolute error from 14.71 bpm (CHRO) and 9.91 (kNN) to 4.51 bpm (ANNMS) and enhances the success-rate-10, the probability in which the absolute error is below 10 bpm, from 44.1% (CHRO) and 56.3% (kNN) to 91.5% (ANNMS).

A novel Adaptive Genetic Neural Network (AGNN) model for recommender systems using modified k-means clustering approach

The Recommender System (RS) plays an important role in information retrieval techniques in a bid to handle massive online data effectively. It gives suggestions on items/services to the target online user to ensure correct decisions quickly and easily. Collaborative Filtering (CF) is a key approach in RS providing a recommendation to the target online user, based on a rating similarity among users. Unsupervised clustering approach is a model-based CF, which is preferred as it ensures simple and effective recommendation. Such CFs suffer from a high error rate and needs additional iterations for convergence. This paper proposes a Modified k-means clustering approach to eliminate the above mentioned issues to provide well-framed clusters. The novel supervised Adaptive Genetic Neural Network (AGNN) method is proposed to locate the most favored data points in a cluster to deliver effective recommendations. The performance of the proposed RS is measured by conducting an experimental analysis on benchmark MovieLens and Netflix datasets. Results are compared with state-of-the-art methods namely Artificial Neural Network (ANN) and Fuzzy based RS models to show the effectiveness of the proposed AGNN method.

Evaluation of Scale Effect of Fragmented Agricultural Land Transfer Based on Neural Network

The artificial neural network consists of a large number of neurons interconnected to each other, which simulates the information process mode of human brains. This is a complex network system that carries out parallel processing of information and nonlinear transformation. The agricultural land in China is characterized by fine fragmentation, and people-land relation is tense. In order to promote the intensive development of agriculture in China, this paper takes 31 provinces and cities as research objects and proposes an evaluation model based on adaptive fuzzy neural network (ANFIS) for the scale effect of fragmented agricultural land transfer to overcome the subjectivity of traditional evaluation methods. In order to improve the accuracy of the evaluation, the combination of qualitative and quantitative methods is used to screen the indicators, and the subtraction clustering method is used to obtain the initial fuzzy rules of the adaptive fuzzy neural network. Then, the adaptive fuzzy neural network model can be trained and tested and the convergence results of the model are good. The results of model analysis show that under different resource endowments and production conditions, the scale effect of land transfer in different regions presents significant differences. On this basis, it is proposed that the intensification process of the agricultural land use cannot be blindly promoted, and the land transfer should follow the principle of moderation. Appropriate agricultural production models should be formulated based on different geographical environments.

Ensemble empirical mode decomposition based adaptive wavelet neural network method for wind speed prediction

Wind energy is one of the emerging sustainable sources of electricity. Wind is intermittent in nature. The typical grid operation of wind energy is complex. The significance of wind energy generation and integration with the grid is increasing day by day. An accurate wind speed forecasting method will help the utility planners and operators to meet the balance of supply and demand by generating wind energy. In this paper, a statistical-based wind speed prediction is implemented without utilizing the numerical weather prediction inputs. This analytical study proposes a hybrid short-term prediction approach that can successfully preprocess the original wind speed data to enhance the forecasting accuracy. The most efficient signal decomposition algorithm, Ensemble Empirical Mode Decomposition is used for preprocessing. This ensemble empirical mode decomposition technique decomposes the original wind speed data. Each decomposed signal is regressed to forecast the future wind speed value by utilizing the Adaptive Wavelet Neural Network model. The proposed hybrid approach is subsequently investigated with respect to the wind farm of South India. The results from a real-world case study in India are reported along with comprehensive comparison. The prediction performance delivered high accuracy, less uncertainty and low computational burden in the forecasts attained. The developed hybrid model outperforms the six other benchmark models such as persistence method, back propagation neural network, radial basis function neural network, Elman neural network, Gaussian regression neural network, and wavelet neural network.

Understanding Prediabetes in a Medicare Advantage Population Using Data Adaptive Techniques.

The objective was to identify individuals with undiagnosed prediabetes from administrative data using adaptive techniques. The data source was a national Medicare Advantage Prescription Drug (MAPD) plan administrative data set. A retrospective, cross-sectional study developed and evaluated data adaptive logistic regression, decision tree, neural network, and ensemble predictive models for metabolic syndrome and prediabetes using 3 mutually exclusive cohorts (N = 279,903). The misclassification rate (MCR), average squared error (ASE), c-statistics, sensitivity (SN), and false positive (FP) rates were compared to select the final predictive models. MAPD individuals with continuous enrollment from 2013 to 2014 were included. Metabolic syndrome and prediabetes were defined using clinical guidelines, diagnosis, and laboratory data. A total of 512 variables identified through subject matter expertise in addition to utilizing all data available were evaluated for the modeling. The ensemble model demonstrated better discrimination (c-statistics, MCR, and ASE of 0.83, 0.24, and 0.16, respectively), high SN, and low FP rate in predicting metabolic syndrome than the individual data adaptive modeling techniques. Logistic regression demonstrated better discrimination (c-statistics, MCR, and ASE of 0.67, 0.13, and 0.11 respectively), high SN, and low FP rate in predicting prediabetes than the other adaptive modeling techniques or ensemble methods. The scored data predicted prediabetes in 44% of the MAPD population, which is comparable to 2005-2006 National Health and Nutrition Examination Survey prediabetes rates of 41%. The logistic regression model demonstrated good performance in predicting undiagnosed prediabetes in MAPD individuals.

Intelligent Manufacturing: TCAD-Assisted Adaptive Weighting Neural Networks

Using machine intelligence on device and process performance prediction is an emerging methodology in the IC industry. While semiconductor technology computer-aided design (TCAD) has been researched and developed for over 30 years, it should contribute to or be used in conjunction with machine learning algorithms in solution finding procedure. Here, we propose an adaptive weighting neural network (AWNN) model that combines the advantages of statistical the machine learning model and the physical TCAD model. Using aspect ratio dependent etching as an example, our proposed AWNN outperforms conventional artificial neural network in terms of mean square errors in the test set where 5-10 times reduction is observed. The effectiveness of the TCAD AWNN model can be especially effective in the case of sampling over a vast sample space since the under-sampling problem can be compensated by the TCAD model. The large and nearly unbounded sample space is very common in IC technology, where cascaded and repeated process steps exist (~150 process steps and ~20 masks for 90-nm CMOS process).