Unsupervised Neural Network Method Research Articles

Gas channels and chimneys prediction using artificial neural networks and multi-seismic attributes, offshore West Nile Delta, Egypt

Machine learning techniques combined with multi-seismic attributes and well logs datasets have been successfully used in reducing the risk of drilling operations and petroleum exploration by providing precise petrophysical and seismic information extracted from the hydrocarbon reservoir rocks. For this purpose, Artificial Neural Networks (ANNs) work as a multi-channel processing system with a high degree of interconnection to classify various faces and predict the reservoir properties through the seismic profile by involving multi-seismic attributes and optionally well logs to the inputs. The main aim of this study is to use both supervised and unsupervised neural networks for the first time in the West Delta Deep Marine (WDDM) concession to identify the spatial dimensions of the gas-bearing channels and the detection of gas chimneys across the seismic profiles. We use back-error propagation algorithms of the Multilayer Perceptron (MLP) and self-organizing Unsupervised Vector Quantizer (UVQ) as supervised and unsupervised neural network methods, respectively, to detect the gas zones and channels, and to classify the gas chimneys and non-gas chimneys zones, as well as classification of the seismic reflections and lithologies. The output acquires a detailed analysis of the distribution pattern of gas channels and accurate information to image the gas chimneys. In the current study, the approach adopted is beneficial to image the gas chimneys and channels in different basins in any region of the world with similar geological settings. • Artificial Intelligence (AI) identifies the spatial dimensions of the gas-bearing channels. • Use both Multi-Layer Perceptron (MLP) and self-organizing Unsupervised Vector Quantizer (UVQ) to detect the gas zones and channels, classify the gas chimneys and non-gas chimneys zones. • The role of ANNs techniques combined with multi-seismic attributes and well logs in hydrocarbon exploration.

Leveraging artificial intelligence to analyze the COVID-19 distribution pattern based on socio-economic determinants

The spatialization of socioeconomic data can be used and integrated with other sources of information to reveal valuable insights. Such data can be utilized to infer different variations, such as the dynamics of city dwellers and their spatial and temporal variability. This work focuses on such applications to explore the underlying association between socioeconomic characteristics of different geographical regions in Dublin, Ireland, and the number of confirmed COVID cases in each area. Our aim is to implement a machine learning approach to identify demographic characteristics and spatial patterns. Spatial analysis was used to describe the pattern of interest in electoral divisions (ED), which are the legally defined administrative areas in the Republic of Ireland for which population statistics are published from the census data. We used the most informative variables of the census data to model the number of infected people in different regions at ED level. Seven clusters detected by implementing an unsupervised neural network method. The distribution of people who have contracted the virus was studied.

Enhanced Formation Evaluation in the Permian Basin Using a Novel Field-Scale Workflow Including Wells with Missing Data

Summary Reliable formation evaluation in organic-rich mudrocks requires integrated interpretation of well logs and core measurements. More than 80% of the Permian Basin wells have incomplete data sets, lacking photoelectric factor (PEF) or other logs required for reliable formation evaluation in the presence of complex mineralogy. Hence, we develop a novel workflow to reliably estimate rock properties in wells with incomplete data to enhance reservoir characterization and completion decisions. We propose to use integrated rock classification for enhanced physics-based assessment of rock properties in wells with missing data; combine field-scale geostatistical and machine learning methods to reliably reconstruct missing PEF logs with a confidence interval through a rock-type-based approach, which is a unique contribution of this work; and quantify the uncertainty in estimates of petrophysical properties. We performed a preliminary field-scale formation evaluation on wells with triple-combo logs (more than 70 wells). Next, we performed an initial rock typing and reconstructed the missing PEF logs by combining supervised neural networks with geostatistical analysis on a rock-type basis. We then used an unsupervised neural network method to improve the rock classification based on the updated estimates of petrophysical and compositional properties after PEF reconstruction. The combined rock classification and PEF reconstruction was performed iteratively to improve the multimineral analysis results in all wells with missing data. We successfully applied the new workflow to 20 wells in blind tests. The reconstructed well logs agreed with the actual measurements with relative errors of less than 10%. The new workflow extends the boundaries of reliable formation evaluation, enabling accurate reservoir characterization and completion decisions by enhancing evaluation of wells with missing data. The proposed method can also be applied to wells with other types of missing data.

Mapping of decadal middle Adriatic oceanographic variability and its relation to the BiOS regime

AbstractWe analyzed long‐term time series of temperature, salinity, and dissolved oxygen (DO) concentrations collected along the Palagruža Sill transect (middle Adriatic) between 1952 and 2010. The data have been mostly collected on seasonal basis, allowing for extraction of seasonal signal from the series. By applying Self‐Organizing Maps (SOM) method, a kind of unsupervised neural network method, the processes on a decadal time scale emerged as the most relevant for changes of oceanographic properties in the middle Adriatic area. Sensitivity studies revealed that oceanographic patterns obtained by SOM were not sensitive to shortening of time series, to removal of data from one station or to removal of DO from the analysis. Simultaneous SOM‐based mapping of sea surface heights in the northern Ionian Sea, with these heights serving as a proxy for the Adriatic‐Ionian Bimodal Oscillating System (BiOS), revealed asymmetry between anticyclonic and cyclonic BiOS patterns and correlated the decadal oscillations in the middle Adriatic with the reversals in the BiOS circulation regimes. These reversals are found to either rapidly change oceanographic properties in the middle Adriatic (e.g., during the Eastern Mediterranean Transient) or to change them with a time lag of 2–3 years. The mapped connections may be used for a short‐time (a few years) forecasting of the Adriatic oceanographic properties or for mapping future climate decadal oscillations as seen by ocean climate models.

Modifying Hebbian Network for Text Cipher

The objective of this work is to design and implement a cryptography system that enables the sender to send message through any channel (even if this channel is insecure) and the receiver to decrypt the received message without allowing any intruder to break the system and extracting the secret information. This work modernize the feedforward neural network, so the secret message will be encrypted by unsupervised neural network method to get the cipher text that can be decrypted using the same network to get the original text. The security of any cipher system depends on the security of the related keys (that are used by the encryption and the decryption processes) and their corresponding lengths. In this work, the key is the final weights that are obtained from the learning process within the neural network stage, So the work can be represented as an update or development for using the neural network to enhance the security of text. As a result for a powerful design, the resulted cipher system provides a high degree of security which satisfies the data confidentially which is the main goal of the most cryptography systems.

Modifying Hebbian Network for Text Cipher

The objective of this work is to design and implement a cryptography system that enables the sender to send message through any channel (even if this channel is insecure) and the receiver to decrypt the received message without allowing any intruder to break the system and extracting the secret information. This work modernize the feedforward neural network, so the secret message will be encrypted by unsupervised neural network method to get the cipher text that can be decrypted using the same network to get the original text. The security of any cipher system depends on the security of the related keys (that are used by the encryption and the decryption processes) and their corresponding lengths. In this work, the key is the final weights that are obtained from the learning process within the neural network stage, So the work can be represented as an update or development for using the neural network to enhance the security of text. As a result for a powerful design, the resulted cipher system provides a high degree of security which satisfies the data confidentially which is the main goal of the most cryptography systems.

Hybrid Cipher System using Neural Network

The objective of this work is to design and implement a cryptography system that enables the sender to send message through any channel (even if this channel is insecure) and the receiver to decrypt the received message without allowing any intruder to break the system and extracting the secret information. In this work, we implement an interaction between the feedforward neural network and the stream cipher, so the secret message will be encrypted by unsupervised neural network method in addition to the first encryption process which is performed by the stream cipher method. The security of any cipher system depends on the security of the related keys (that are used by the encryption and the decryption processes) and their corresponding lengths. In our work, there are two types of keys; the first type is the keystream that is adopted by the stream cipher stage with optimal length (length of the keystream greater or equal the message length); and the second key type is the final weights that are obtained from the learning process within the neural network stage, So we can represent our work as an update or development for using the neural network to enhance the security of stream cipher. As a result for a powerful hybrid design, the resulted cipher system provides a high degree of security which satisfies the data confidentially which is the main goal of the most cryptography systems.

Hybrid Cipher System using Neural Network

The objective of this work is to design and implement a cryptography system that enables the sender to send message through any channel (even if this channel is insecure) and the receiver to decrypt the received message without allowing any intruder to break the system and extracting the secret information. In this work, we implement an interaction between the feedforward neural network and the stream cipher, so the secret message will be encrypted by unsupervised neural network method in addition to the first encryption process which is performed by the stream cipher method. The security of any cipher system depends on the security of the related keys (that are used by the encryption and the decryption processes) and their corresponding lengths. In our work, there are two types of keys; the first type is the keystream that is adopted by the stream cipher stage with optimal length (length of the keystream greater or equal the message length); and the second key type is the final weights that are obtained from the learning process within the neural network stage, So we can represent our work as an update or development for using the neural network to enhance the security of stream cipher. As a result for a powerful hybrid design, the resulted cipher system provides a high degree of security which satisfies the data confidentially which is the main goal of the most cryptography systems.

Supervised and unsupervised neural network methods applied to textile quality control based on improved wavelet feature extraction techniques

This paper aims at investigating novel solutions to the problem of textile defect detection from images, that can find applications in building robust quality control vision based systems in textile production. The proposed solutions focus on detecting defects from the textural properties of their corresponding wavelet transformed images. More specifically a novel methodology is investigated for discriminating defects in textile images by applying supervised and unsupervised neural classification techniques, employing multilayer perceptrons (MLP)-trained with the on-line backpropagation algorithm and Kohonen's Self-Organizing Feature Maps (SOFM) respectively. These parallel techniques are applied to innovative wavelet based feature vectors. These vectors are extracted from the wavelet transformed original images using the cooccurrence matrices framework and SVD analysis. The results of the proposed methodology are illustrated in defective textile images where the defective areas are recognized with about 98.5% accuracy.

An Application of Unsupervised Neural Network Methodology Kohonen Topology‐Preserving Mapping to QSAR Analysis

AbstractThe concept and methodology of artificial neural networks is introduced. Like pattern recognition, the techniques can be classified as supervised (requiring a priori knowledge of class membership) and unsupervised (making no assumptions about class membership). An unsupervised neural network method, Kohonen Topology‐Preserving Mapping, is applied to a wide matrix of physicochemical property data for a set of antifilarial antimycin analogues containing structural outliers. Principal component analysis failed to give a good 2D representation of the data set as a whole due to linear constraints in the model which gave undue influence to the outliers. Kohonen mapping compared favourably with non‐linear unsupervised statistical pattern recognition methods for 2D representation of compound similarity and for classification based on antifilarial activity. It may prove a valuable technique for QSAR in situations where a linear method does not model the data well and a high throughput of test compounds is indicated.