Neural Net Technology Research Articles

Применение нейронной сети для управления системой домашней гидропоники

Hydroponics is a method of soilless cultivation of plants. It shortens the vegetation period, reduces the risk of disease and insect infestation, and provides a year-round growing cycle. Hydroponics depends on efficient water management. It is associated with a complex design, operation, and maintenance. Neural networks can control complex technological processes in agriculture. The research objective was to use a neural network to increase the efficiency of a home hydroponics system.
 The study involved a nutrient bed hydroponics setup with ten Lactuca sativa plants. Sensors collected information about the temperature and humidity of air, illumination, and the temperature of the leaf surface. Data processing, neural network training, and microcontroller programming relied on Python 3, PyTorch, and MicroPython.
 The four-layer perceptron, which is a popular control mechanism, turned out to be the most effective neural network architecture. Fewer layers resulted in a high error rate (≥ 5%). When the number of layers was > 4, the error level remained at that of the four-layer experiment (0.2%). Further practical tests showed an increase in energy efficiency by 32.3%, compared to the classical control algorithm at close values of plant transpiration.
 Neural net technology could be integrated into energy-saving residential premises and smart home systems in order to increase the self-sufficiency of hydroponics installations.

Neural Image Classifiers for Historical Building Elements and Typologies

New technologies of machine vision and artificial intelligence (AI) are opening fresh avenues to catalog and compare the entire corpus of built architecture. While neural net technology is rightly embraced as a promising generative paradigm for architecture, it also holds enormous promise for historical work, notably the automatic scanning and organization of as-built imagery and video of buildings and cities. We argue that one may apply AI-driven machine vision tools to scan and classify architectural imagery based on stylistic and morphological considerations. Combined with data science methods, such tools enable a comprehensive view of historic architectural features and types.

High Performance SiGe HBT Performance Variability Learning by Utilizing Neural Networks and Technology Computer Aided Design

Process sensitivity and variation, such as layer thicknesses, etch dimensions and doping levels are all process parameters that should be well understood when assessing their impact on end of process wafer quality and device performance. Improvements in uniformity of electrical device performance is often not taken into major consideration until a certain maturity level of a technology is reached. In this work, use of technology computer aided design (TCAD) and process compact models (PCM) developed from neural networks demonstrate their utility in process- parameter variation understanding in earlier stages of technology development. A TCAD simulation was built and calibrated to match key electrical performance metrics of an experimental high performance SiGe HBT in 130nm BiCMOS technology, which was the device of focus in this study. Figure 1 includes a TCAD cross section of the device1. Neural net techniques, a machine-learning methodology, the development and deployment of which has grown significantly over the last three decades, is used to expand the scope of TCAD to process variability2-3.The aforementioned calibrated TCAD deck was used to generate training data for the neural network process compact model in place of hardware data. Key process features, as listed in table 1 were a few of the process conditions that were varied from nominal values to characterize the variation in resulting electrical performance. Figure 2 includes a 1D schematic view of the epitaxial layers of the HBT, highlighting some of the process features varied in simulation. The neural network was trained using a data analysis suite using three hidden layers with 16 neurons each to get a best fit to the training data input and output values.The foundation of this approach was a TCAD simulation calibrated to match performance of the nominal HBT device of focus. Key AC and Gummel performance parameters had good agreement with hardware data as illustrated in Figure 3a and b. Thousands of individual TCAD simulations with combinations of varied process feature values were then executed to generate neural network training data in lieu of hardware data- utilizing reported standard deviations of each process parameter. The trends and sensitivity of this data was then reflected in the process compact model, which would provide results for large-scale calculation and analysis. Beta was the electrical parameter of focus in this study, where its changes in process variation (mean and standard deviation) were predicted by the trained PCM. The agreement with hardware electrical parameter Beta with respect to Boron dose in Figure 4 demonstrates how use of calibrated TCAD as a basis can empower prediction of large-scale hardware results without the extent of hardware expense. Use of this tool can be extended to simulate and analyze larger populations than what would be feasible for hardware (in several thousands) to predict the effect of process input changes on uniformity of electrical parameters. Figure 5 demonstrates how the changes in standard deviation of Beta can be predicted as a function of Boron dose variation-through generation of large simulated populations by the PCM.The resulting work on a state-of-the-art SiGe BiCMOS technology demonstrates how specific simulation tools, like TCAD-trained process compact models can be leveraged for enhanced process understanding and improvement. The data generated from these tools can supplement or replace hardware and corresponding data, saving in development costs and associated hours in manpower. They can also enable earlier and more agile decision-making about key production metrics throughout the technology development process. Calibrated TCAD infrastructure, coupled with neural-net technology have potential to fulfill pivotal roles through the timelines of technology development, rather than just being utilized when a technology is mature.

Convolution Neural Network: A Shallow Dive in to Deep Neural Net Technology

It is always beneficial to reassess the previously done work to create interest and develop understanding about the subject in importance. In computer vision, to perform the task of feature extraction, classification or segmentation, measurement and assessment of image structures (medical images, natural images etc.) is to be done very efficiently. In the field of image processing numerous techniques are available, but it is very difficult to perform these tasks due to noise and other variable artifacts. Various Deep machine learning algorithms are used to perform complex task of recognition and computer vision. Recently Convolutional Neural Networks (CNNs-back bone of numerous deep learning algorithms) have shown state of the art performance in high level computer vision tasks, such as object detection, object recognition, classification, machine translation, semantic segmentation, speech recognition, scene labelling, medical imaging, robotics and control, , natural language processing (NLP), bio-informatics, cybersecurity, and many others. Convolution neural networks is the attempt to combine mathematics to computer science with icing of biology on it. CNNs work in two parts. The first part is mathematics that supports feature extraction and second part is about classification and prediction at pixel level. This review is intended for those who want to grab the complete knowledge about CNN, their development form ancient age to modern state of art system of deep learning system. This review paper is organized in three steps: in the first step introduction about the concept is given along with necessary background information. In the second step other highlights and related work proposed by various authors is explained. Third step is the complete layer wise architecture of convolution networks. The last section is followed by detailed discussion on improvements, and challenges on these deep learning techniques. Most papers consider for this review are later than 2012 from when the history of convolution neural networks and deep learning begins

Validating Black Box Neural Networks

Machine learning, driven by recent advances in neural net technology, holds much promise, but how to validate any particular model? This note looks at why validation is necessary, and describes some practical techniques for doing so. The example scenario (from a Health Insurance survey, N = 29,145) is a sentiment analysis of verbatims arising from the Net Promoter Score expatiate ‘Why that rating?’ We show that respondents’ ratings can be predicted to a useful level of accuracy, and that the model can be validated by reference to external benchmarks.

Neural-net modeling for direct and inverse problems of shell theory

Effectiveness of the use of neural-net technology for the solving of shell theory problems is shown. Some results of neural-net interpolation and extrapolation for direct and inverse problems are discussed. Exact accuracy of neural-net solving opens wide latitude for shell constructions engineering design and optimization.

Joint Optimization of Spectro-Temporal Features and Deep Neural Nets for Robust Automatic Speech Recognition

In speech recognition, feature extraction and acoustical model training are traditionally done in two separate steps. Here, instead, we use a framework that combines spectro-temporal feature extraction and the training of neural network based acoustic models into a single process. We found earlier that this approach can be successfully applied for the recognition of speech. In this paper, we propose two further improvements to our method based on recent advances in neural net technology and extend our evaluation to speech contaminated with new types of noise. By repeating our experiments on TIMIT phone recognition tasks using clean and noise contaminated speech, we can compare the recognition performance of the original framework with our new, modified framework. The results indicate that both these modifications significantly improve the recognition performance of our framework. Moreover, we will show that these modifications allow us to achieve a substantially better performance than what we got earlier.

Data fusion control and guidance of surface-to-air missile under the complex circumstance based on neural-net technology

Under the complicated electromagnetism circumstance, the model of data fusion control and guidance of surface-to-air missile weapon systems is established. Such ways and theories as Elman-NN, radar tracking and filter's data fusion net based on the group method for data-processing (GMRDF) are applied to constructing the model of data fusion. The highly reliable state estimation of the tracking targets and the improvement in accuracy of control and guidance are obtained. The purpose is optimization design of data fusion control and guidance of surface-to-air missile weapon systems and improving the fighting effectiveness of surface-to-air missile weapon systems.

Hierarchical neural net technology for molecular staging with clinical data

Hierarchical neural net technology for molecular staging with clinical data

ABSTRACT: Application of neural net technology and climate-stratigraphy in stratigraphic correlations

ABSTRACT: Application of neural net technology and climate-stratigraphy in stratigraphic correlations

Artificial neural networks and their business applications

Artificial neural networks are increasingly popular in today's business fields. They have been hailed as the greatest technological advance since the invention of transistors. The purpose of this paper is to answer two of the most frequently asked questions: “What are neural networks?” “Why are they so popular in today's business fields?” The paper reviews the common characteristics of neural networks and discusses the feasibility of neural-net applications in business fields. It then presents four actual application cases and identifies the limitations of the current neural-net technology.

Massively parallel structural design using stochastic optimization and mixed neuralnet/finite element analysis methods

The title study is performed on the massively parallel processing (MPP) environment of Connection Machine (CM) computers using truss structural sizing design problems as example design problems. In this design optimization procedure, only the displacement solution is replaced by that based on neural net technology (under a given set of cross-sectional size parameters (e.g., areas) in the MPP finite element structural reanalysis). This structural reanalysis procedure, together with a vastly improved and parallelized version of the integral global optimization (IGO) stochastic algorithm, IIGO, forms the present MPP structural design methodology. In addition, a procedure to correct the final optimal design for constraint violation or too-conservatively satisfied constraint condition caused by inaccuracy of the NN (neural network) analysis model is also formulated. Evaluation of the numerical performance of the developed computational algorithm set, capability, and strategy is made, primarily on the Connection Machine CM-2 model computer by performing three neural-network-based truss structural reanalysis/minimum weight design problems.

A neuro-fuzzy system for X-ray spectra interpretation

A decision scheme for the interpretation of spectra from wavelength dispersive X-ray fluorescence spectrometry is described that encompasses elements from three areas of artificial intelligence: fuzzy logic, rule based expert systems and neural net technology. After transforming the recorded spectra to line spectra by appropriate background correction a reasoning scheme is applied that takes into account not only the observed spectra, but also the recording conditions and prior spectroscopic information regarding the relative emission probabilities and the usefulness of the different lines for the purpose of element identification. The latter is done on the basis of a previously described scheme to compute conditional a posteriori Bayes probabilities for a “mean matrix”. These different pieces of information are then assembled into a battery of fuzzy rules. The importance of the rules as well as the importance of the X-ray lines is determined in a training process, similar to the one in a feedforward back-propagation network. To further stabilize the results this network is pruned in a second training cycle. This, however, had little effect on the quality of interpretation. The advantages of this approach to the interpretation of X-ray spectra over older ones are numerous: the system adapts itself to better interpret spectra that are of greater importance to a laboratory as these are better represented in the training set; the fuzzy logic is capable of working with incomplete and uncertain knowledge, and the neural network results based on these fuzzy rules is readily interpretable by the X-ray spectroscopist as every rule can be expressed also in natural language as in any classical rule based system.

Using Artificial Neural Nets To Identify the Well-Test Interpretation Model

Summary This paper presents a new approach, based on artificial neural net technology, to identify a preliminary well-test interpretation model from derivative plot data. Artificial neural nets can identify patterns from incomplete and distorted data and also eliminate the need for elaborate data preparation, such as smoothing.

Neural network based calculation of short-circuit currents in three-phase systems

In this paper we present a new method for short-circuit studies in three-phase systems based on the use of supervised learning neural net technology and the adaptive pattern recognition concept. Neural nets are used to assess the essential characteristics of short-circuit currents. The main motivation is to exploit generalization capabilities of neural nets to interpolate between training data, and thus to allow fast and direct assessment of short-circuit currents in cases of changing network topology and parameters. We present the results obtained in computer simulations using the example of a power system.

Neural-net based unstable machine identification using individual energy functions

The identification of the mode of instability plays an essential role in generating principal energy boundary hypersurfaces. We present a new method for unstable machine identification based on the use of supervised learning neural-net technology, and the adaptive pattern recognition concept. It is shown that using individual energy functions as pattern features, appropriately trained neural-nets can retrieve the reliable characterization of the transient process including critical clearing time parameter, mode of instability and energy margins. Generalization capabilities of the nueral-net processing allow for these assessments to be made independently of load levels. The results obtained from computer simulations are presented using the New England power system, as an example.

Applying neural net technology for multi-objective land use planning

Neural nets have been studied for many years in an attempt to achieve human-like performance in fields where problems are unlikely to be solved efficiently by using traditional methods. Yet there have been very few practical applications of neural nets particularly in resource management. In this study, the neural net technology is adopted as a decision support system for land use planning. Computer simulation software for such a net is developed and an empirical analysis is undertaken for the Peace River Region of British Columbia in Canada to deal with problems related to multi-objective land planning. This neural net model incorporates biophysical and socio-economic data from various sources and reflects the interactions between different land use objectives. Given a set of land use options and their geographical requirements, the neural net identifies the desired land use options among the regions. The neural net model applied in this study provides an introduction to a useful computer technology for natural resource management.