Search

Based on our previous work, an improved algorithm for splitting raw data into training and test subsamples was developed to predict the spatial distribution of chemical elements in the surface soil layer using artificial neural networks (ANNs). The algorithm takes into account various factors affecting the distribution of the modeled variable. The data was obtained during soil screening on the urbanized area of the Noyabrsk city. [ABSTRACT FROM AUTHOR]

The paper describes the application of the rapid serial visual presentation (SRVP) method to study the dynamics of recognition time of visual stimuli by an untrained observer. The method is implemented on the basis of a specially developed program presenting stimulus material on a computer screen according to a given algorithm. Raster monochrome images of two letters of the Russian alphabet, depicted in two significantly different styles, acted as stimulus material. Presentation was carried out in a specialized viewing camera. When evaluating the results of the experiment, ruggedness was taken into account as a quantitative measure of the lettering. In the experiment, an increase in the recognition time of all types of stimuli with an increase in the number of presentations was found. Using the program greatly facilitated the experiment. [ABSTRACT FROM AUTHOR]

An algorithm based on predicting the residuals of the nonlinear autoregressive neural network model with external input (NARX), which can improve the prediction accuracy, was proposed. Data of the concentration of one of the main greenhouse gases methane (CH4) on the Arctic Island of Belyy, Russia, were used for prediction. A time interval, which was characterized by high daily fluctuations in the CH4 concentration was selected. The forecast accuracy was determined by the mean absolute error (MAE), root mean squared error (RMSE) and root mean squared relative error (RMSRE) errors. The use of the algorithm allowed to increase the forecast accuracy from 11% for RMSE to 20% for RMSRE. [ABSTRACT FROM AUTHOR]

When studying the processes associated with global warming, forecasts of time series are very important. The present study used data of the concentration of greenhouse gas methane in the surface layer of atmospheric air on the Arctic island Belyi, Russia. For the work, a time interval of 170 hours was chosen. For the modelling, a model based on a nonlinear autoregressive neural network with an external input (NARX) was used. As a training algorithm three types were applied: Levenberg-Marquart (LM), LM with Bayesian regularization (BR), and a gradient descent with a speed parameter setting (GDA). Methane concentrations corresponding to the first 150 hours of the interval used to train the NARX network, then the concentrations were predicted for the next 20 hours. The model using the training algorithm LM was the most accurate. [ABSTRACT FROM AUTHOR]

The paper is present a comparison of modern approaches for predicting the spatial distribution in the upper soil layer of a chemical element chromium (Cr), which had spots of anomalously high concentration in the investigated region. The distribution of a normally distributed element copper (Cu) was also predicted. The data were obtained as a result of soil screening in the city of Tarko-Sale, Russia. Models based on artificial neural networks (multilayer perceptron MLP), random forests (RF), and also a model based on a random forest in which MLP used as a tree - a random perceptron forest (RMLPF) - were considered. The models were implemented in MATLAB. Approaches using artificial neural networks (MLP and RMLPF) were significantly more accurate for anomalously distributed Cr. Models based on RF algorithms proved to be more accurate for normally distributed copper. In general, the proposed model RMLPF was the most universal and accurate. [ABSTRACT FROM AUTHOR]