Techniques for Selecting the Optimal Parameters of One-Class Support Vector Machine Classifier for Reduced Samples

Usually, the One-Class Support Vector Machine (OC-SVM) requires a large dataset for modeling effectively the target class independently to other classes. For finding the OC-SVM model, the available dataset is subdivided into two subsets namely training and validation, which are used for training and validating the optimal parameters. This approach is effective when a large dataset is available. However, when training samples are reduced, parameters of the OC-SVM are difficult to find in absence of the validation subset. Hence, this paper proposes various techniques for selecting the optimal parameters using only a training subset. The experimental evaluation conducted on several real-world benchmarks proves the effective use of the new selection parameter techniques for validating the model of OC-SVM classifiers versus the standard validation techniques

Mask detection as Covid combat using Transfer learning and Haar Cascade

The World is going through a pandemic due to the rapid transmission of COVID-19. According to the several guidelines issued by WHO (World Health Organization), wearing a mask is the most effective preventive measure in public/crowded places. We hope for the future social health and safety of the people around the world with this project. To detect the people who are not following the COVID-19 guidelines in public/crowded areas a convolutional neural network under the framework of the TensorFlow VGG-19 algorithm is proposed which has trained and tested a collection of more than 1350 images. One flat layer and two FC layers with reduced parameters are optimized from three FC layers. The 2-label softmax classifier replaced the softmax classification layer of the original model. Our experimental results show a training accuracy of 99.73% and an accuracy of 98.78% during testing. Keywords: Transfer learning, covid19, mask detection, artificial intelligence, coronavirus

Stability Metric Based on Sensitivity Analysis Applied to Electrical Repowering System

Stability metrics are used to quantify a system’s ability to maintain equilibrium under disturbances. We did not identify the proposition of a stability metric using sensitivity analysis within the literature. This work proposes a system stability metric and its application to an electrical repowering system. The methodology for applying the proposed metric comprises: (i) system parameters sensitivity analysis and spider diagram construction, (ii) determining the array containing the line segments inclination angles of each spider diagram curve, and (iii) stability calculation using the array mean and maximum inclination value of a line segment. After simulating the model built for the electrical repowering system and applying the methodology, we obtain results regarding the sensitivity indices and stability values of system inputs relative to their outputs, considering the original system and with reduced parameters. Using the stability study, it was possible to determine different stability categories for the system parameters, which indicates the need for different analysis levels.

Physical and Chemical Fundamentals of Building the Structure of Dispersed Filled Polymer Composite Materials and Nanocomposites

Questions of the construction of dispersed structures of polymer composite materials using a generalized model of dispersed filled polymer composite materials (DFPCM) are studied. Using the parameter of maximum proportion of filler (φm) allows you to take into account the size, shape, and distribution of part of the dispersed filler at the same time. The transition to generalized and specified parameters when describing the structure of the DFPCM leads to the possibility of highlighting the optimal criteria for obtaining systems with the highest strength characteristics. The transition to generalized and reduced parameters when describing the structure of DFPCM leads to the possibility of selecting optimal criteria that ensure obtaining systems with the necessary level of rheological, electrochemical, physico-mechanical and other characteristics.

Investigation of the Effect of Combined Binder Based on Portland Cement and Lime on the Properties of Non-Autoclave Silicate Materials Modified with Synthetic Crystalline Filler

Among the wide variety of currently used wall building materials and products, it is possible to single out the autoclave-hardened silicate products. To obtain silicate materials of autoclave hardening, lime-silica binders are mainly used. The hardening process of such a binder is carried out in an environment of water vapor at high temperature and pressure. It is relevant to use a certain type of silicate materials in the technology with raw materials that provide hardening with reduced parameters of hydrothermal synthesis, which will make it possible to obtain wall silicate materials in non-autoclave conditions. This can be done through the use of clay rocks of the mineral formation unfinished stage. In the course of the research, the effect of a combined binder based on Portland cement and lime on the properties of non-autoclave silicate materials modified with a synthetic crystalline filler was studied. It has been established that the addition of a synthetic crystalline filler represented by artificial calcium hydro-silicates makes it possible to increase the presence of a crystalline phase formed due to the clinker minerals hydration in the early stages of hardening, and, as a consequence, to increase the operational properties of the resulting composites. The addition of lime to the raw mixture will additionally compensate for the lack of calcium ions during the entire hardening process of non-autoclave silicate composites based on aluminosilicate raw materials and Portland cement.

Generalized Hurst Hypothesis: Description of Time-Series in Communication Systems

In this paper, we focus on the generalization of the Hurst empirical law and suggest a set of reduced parameters for quantitative description of long-time series. These series are usually considered as a specific response of a complex system (economic, geophysical, electromagnetic and other systems), where successive fixations of external factors become impossible. We consider applying generalized Hurst laws to obtain a new set of reduced parameters in data associated with communication systems. We analyze three hypotheses. The first one contains one power-law exponent. The second one incorporates two power-law exponents, which are in many cases complex-conjugated. The third hypothesis has three power-law exponents, two of which are complex-conjugated as well. These hypotheses describe with acceptable accuracy (relative error does not exceed 2%) a wide set of trendless sequences (TLS) associated with radiometric measurements. Generalized Hurst laws operate with R/S curves not only in the asymptotic region, but in the entire domain. The fitting parameters can be used as the reduced parameters for the description of the given data. The paper demonstrates that this general approach can also be applied to other TLS.

Depth-Wise Separable Convolution Neural Network with Residual Connection for Hyperspectral Image Classification

The neural network-based hyperspectral images (HSI) classification model has a deep structure, which leads to the increase of training parameters, long training time, and excessive computational cost. The deepened network models are likely to cause the problem of gradient disappearance, which limits further improvement for its classification accuracy. To this end, a residual unit with fewer training parameters were constructed by combining the residual connection with the depth-wise separable convolution. With the increased depth of the network, the number of output channels of each residual unit increases linearly with a small amplitude. The deepened network can continuously extract the spectral and spatial features while building a cone network structure by stacking the residual units. At the end of executing the model, a 1 × 1 convolution layer combined with a global average pooling layer can be used to replace the traditional fully connected layer to complete the classification with reduced parameters needed in the network. Experiments were conducted on three benchmark HSI datasets: Indian Pines, Pavia University, and Kennedy Space Center. The overall classification accuracy was 98.85%, 99.58%, and 99.96% respectively. Compared with other classification methods, the proposed network model guarantees a higher classification accuracy while spending less time on training and testing sample sites.