Multiple Layer Perceptron Neural Network Research Articles

A scanner based neural network technique for color matching of dyed cotton with reactive dye

Conventional theory for color matching is Kubelka-Munk, but it fails in some situations. New intelligent procedures such as neural networks could learn the behavior of a complex system and produce accurate prediction. This paper investigates the ability of MLP (multiple layer perceptron) neural network for color matching of cotton fabric. Three reactive dyes, namely Levafix Red CA, Levafix Yellow CA and Levafix Blue CA were used for experiments. The dyed samples were scanned and L * a * b * histogram were extracted. Different neural networks were trained and tested using L * a * b * histogram of fabric’s images and also L * a * b * values (D65, 10°) of fabrics. The results were encouraging. For neural networks including the L * a * b * histogram in input vector, colorants and their concentration were predicted with a mean square error (MSE) less than 10−5 and an average value of color difference (CMC (1:2)) less than 1.5 for approximately 80 % of testing data.

Macrocell path loss prediction using artificial intelligence techniques

The prediction of propagation loss is a practical non-linear function approximation problem which linear regression or auto-regression models are limited in their ability to handle. However, some computational Intelligence techniques such as artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems (ANFISs) have been shown to have great ability to handle non-linear function approximation and prediction problems. In this study, the multiple layer perceptron neural network (MLP-NN), radial basis function neural network (RBF-NN) and an ANFIS network were trained using actual signal strength measurement taken at certain suburban areas of Bauchi metropolis, Nigeria. The trained networks were then used to predict propagation losses at the stated areas under differing conditions. The predictions were compared with the prediction accuracy of the popular Hata model. It was observed that ANFIS model gave a better fit in all cases having higher R2 values in each case and on average is more robust than MLP and RBF models as it generalises better to a different data.

Abiotic influences on dissolved oxygen in a riverine environment

Dissolved oxygen (DO) is an important parameter in riverine health. Periphyton and/or macrophytes are frequently the drivers behind the fluctuation of DO levels in aquatic environments; however, the effects of abiotic and biotic factors on biomass and in turn on DO may be variable from river to river. The objective of this paper is to understand which factors govern DO levels in terms of daily minimum DO ( DO min) and daily DO variation (Δ DO) in a major wastewater-impacted river using statistical data analysis and modeling. Both climatic conditions (reflected in water temperature) and hydrometric conditions (flow) were major factors influencing DO min and Δ DO. The effect of flow on DO min and Δ DO was discontinuous, depending on the flow magnitude. Nutrient loading from wastewater effluent was not identified to have a significant impact on a daily basis; however, their role over large time scales is unclear. In the data-driven modeling approaches, the non-linear approach using multiple-layer perceptron neural network, which has very flexible architecture, was superior to the linear approach used (multiple linear regression). Although nutrients are likely related to the DO min and Δ DO, flow and temperature were sufficient to obtain robust prediction in the data-driven modeling. This approach is useful to model and understand complicated processes when their governing mechanisms are not well presented in conceptual- or physically based models.

Epilepsy seizure detection using eigen-system spectral estimation and Multiple Layer Perceptron neural network

In this paper, a new approach based on eigen-systems pseudo-spectral estimation methods, namely Eigenvector ( EV) and MUSIC, and Multiple Layer Perceptron (MLP) neural network is introduced. In this approach, the calculated EEG (electroencephalogram) spectrum is divided into smaller frequency sub-bands. Then, a set of features, { maximum, entropy, average, standard deviation, mobility}, are extracted from these sub-bands. Next, incorporating a set of the EEG time domain features { standard deviation, complexity measure} with the spectral feature set, a feature vector is formed. The feature vector is then fetched into a MLP neural network to classify the signal into the following three states: normal ( healthy), epileptic patient signal in a seizure-free interval ( inter-ictal), and epileptic patient signal in a full seizure interval ( ictal). The experimental results show that the classification of the EEG signals maybe achieved with approximately 97.5% accuracy and the variance of 0.095% using an available public EEG signals database. The results are among the best reported methods for classifying the three states aforementioned. This is a high speed with high accuracy as well as low misclassifying rate method so it can make the practical and real-time detection of this chronic disease feasible.

A Signal Processing Module for the Analysis of Heart Sounds and Heart Murmurs

In this paper a Signal Processing Module (SPM) for the computer-aided analysis of heart sounds has been developed. The module reveals important information of cardiovascular disorders and can assist general physician to come up with more accurate and reliable diagnosis at early stages. It can overcome the deficiency of expert doctors in rural as well as urban clinics and hospitals. The module has five main blocks: Data Acquisition & Pre-processing, Segmentation, Feature Extraction, Murmur Detection and Murmur Classification. The heart sounds are first acquired using an electronic stethoscope which has the capability of transferring these signals to the near by workstation using wireless media. Then the signals are segmented into individual cycles as well as individual components using the spectral analysis of heart without using any reference signal like ECG. Then the features are extracted from the individual components using Spectrogram and are used as an input to a MLP (Multiple Layer Perceptron) Neural Network that is trained to detect the presence of heart murmurs. Once the murmur is detected they are classified into seven classes depending on their timing within the cardiac cycle using Smoothed Pseudo Wigner-Ville distribution. The module has been tested with real heart sounds from 40 patients and has proved to be quite efficient and robust while dealing with a large variety of pathological conditions.

Spatial and temporal multiyear sea ice distributions in the Arctic: A neural network analysis of SSM/I data, 1988–2001

Arctic multiyear sea ice concentration maps for January 1988–2001 were generated from SSM/I brightness temperatures (19H, 19V, and 37V) using modified multiple layer perceptron neural networks. Learning data for the neural networks were extracted from ice maps derived from Okean and ERS satellite imagery to capitalize on the stability of active radar multiyear ice signatures. Evaluations of three learning algorithms and several topologies indicated that networks constructed with error back propagation learning and 3‐20‐1 topology produced the most consistent and physically plausible results. Operational neural networks were developed specifically with January learning data, and then used to estimate daily multiyear ice concentrations from daily‐averaged SSM/I brightness temperatures during January. Monthly mean maps were produced for analysis by averaging the respective daily estimates. The 14‐year series of January multiyear ice distributions revealed dense and persistent cover in the central Arctic surrounded by expansive regions of highly fluctuating interannual cover. Estimates of total multiyear ice area by the neural network were intermediate to those of other passive microwave algorithms, but annual fluctuations and trends were similar among all algorithms. When compared to Radarsat estimates of multiyear ice concentration in the Beaufort and Chukchi Seas (1997–1999), average discrepancies were small (0.9–2.5%) and spatial coherency was reasonable, indicating the neural network's Okean and ERS learning data facilitated passive microwave inversion that emulated backscatter signatures. During 1988–2001, total January multiyear ice area declined at a significant linear rate of −54.3 × 103 km2 yr−1 (−1.4% yr−1). The most persistent and extensive decline in multiyear ice concentration (−3.3% yr−1) occurred in the southern Beaufort and Chukchi Seas. In autumn 1996, a large multiyear ice recruitment of over 106 km2 (mostly in the Siberian Arctic) fully replenished the previous 8‐year decline in total area, but it was followed by an accelerated and compensatory decline during the subsequent 4 years. Seventy‐five percent of the interannual variation in January multiyear sea ice area was explained by linear regression on two atmospheric parameters: the previous winter's (JFM) Arctic Oscillation index as a proxy to melt duration and the previous year's average sea level pressure gradient across the Fram Strait as a proxy to annual ice export. Consecutive year changes (1994–2001) in January multiyear ice volume were significantly correlated with duration of the intervening melt season (R2 = 0.73, −80.0 km3 d−1), emphasizing a large thermodynamic influence on the Arctic's mass sea ice balance during summers with anomalous melt durations.

Sensor Fusion for Cutting Tool State Identification in Metal Turning Through the Application of Perceptron Neural Networks

The demand on better returns by investors and the quest for continued customer satisfaction through assured quality products, have culminated in a rapid increase in factory automation processes within the last 20 years. This paper outlines an investigative study, where Artificial Neural Networks are applied to automatically detect the state of a cutting tool. The ensuing wear on the tool is chosen as the parameter to be detected, and process parameters from the cutting environment are measured on-line, fused through the network and its output gives an indication of tool wear level. The implemented Single and Multiple Layer Perceptron Neural Networks achieved success accuracy rates of > 75% and > 95% respectively.