Search results for: stochastic approximation gradient

Authors: Timir Baran Roy, Luis Guerreiro, Ashutosh Bagchi

Abstract:

Dynamic study is important in order to design, repair and rehabilitation of structures. It has played an important role in the behavior characterization of structures; such as bridges, dams, high-rise buildings etc. There had been a substantial development in this area over the last few decades, especially in the field of dynamic identification techniques of structural systems. Frequency Domain Decomposition (FDD) and Time Domain Decomposition are most commonly used methods to identify modal parameters; such as natural frequency, modal damping, and mode shape. The focus of the present research is to study the dynamic characteristics of typical timber masonry walls commonly used in Portugal. For that purpose, a multi-storey structural prototypes of such walls have been tested on a seismic shake table at the National Laboratory for Civil Engineering, Portugal (LNEC). Signal processing has been performed of the output response, which is collected from the shaking table experiment of the prototype using accelerometers. In the present work signal processing of the output response, based on the input response has been done in two ways: FDD and Stochastic Subspace Identification (SSI). In order to estimate the values of the modal parameters, algorithms for FDD are formulated, and parametric functions for the SSI are computed. Finally, estimated values from both the methods are compared to measure the accuracy of both the techniques.

Keywords: frequency domain decomposition (fdd), modal parameters, signal processing, stochastic subspace identification (ssi), time domain decomposition

Procedia PDF Downloads 246

Authors: G. Shahzadi, A. Soulaimani

Abstract:

This research work aims to identify the physical parameters of the constitutive soil model in the design of a rockfill dam by inverse analysis. The best parameters of the constitutive soil model, are those that minimize the objective function, defined as the difference between the measured and numerical results. The Finite Element code (Plaxis) has been utilized for numerical simulation. Polynomial and neural network-based response surfaces have been generated to analyze the relationship between soil parameters and displacements. The performance of surrogate models has been analyzed and compared by evaluating the root mean square error. A comparative study has been done based on objective functions and optimization techniques. Objective functions are categorized by considering measured data with and without uncertainty in instruments, defined by the least square method, which estimates the norm between the predicted displacements and the measured values. Hydro Quebec provided data sets for the measured values of the Romaine-2 dam. Stochastic optimization, an approach that can overcome local minima, and solve non-convex and non-differentiable problems with ease, is used to obtain an optimum value. Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Differential Evolution (DE) are compared for the minimization problem, although all these techniques take time to converge to an optimum value; however, PSO provided the better convergence and best soil parameters. Overall, parameter identification analysis could be effectively used for the rockfill dam application and has the potential to become a valuable tool for geotechnical engineers for assessing dam performance and dam safety.

Keywords: Rockfill dam, parameter identification, stochastic analysis, regression, PLAXIS

Procedia PDF Downloads 122

Authors: Xiaolin Zeng, Lei Cheng, Zhirong Li, Xueping Liu

Abstract:

Aiming at the problem of low accuracy and susceptibility to cultural relic diseases in the generation of high-resolution archaeology drawings by current image generation algorithms, an archaeology drawings generation algorithm based on a conditional generative adversarial network is proposed. An attention mechanism is added into the high-resolution image generation network as the backbone network, which enhances the line feature extraction capability and improves the accuracy of line drawing generation. A dual-branch parallel architecture consisting of two backbone networks is implemented, where the semantic translation branch extracts semantic features from orthophotographs of cultural relics, and the gradient screening branch extracts effective gradient features. Finally, the fusion fine-tuning module combines these two types of features to achieve the generation of high-quality and high-resolution archaeology drawings. Experimental results on the self-constructed archaeology drawings dataset of grotto temple statues show that the proposed algorithm outperforms current mainstream image generation algorithms in terms of pixel accuracy (PA), structural similarity (SSIM), and peak signal-to-noise ratio (PSNR) and can be used to assist in drawing archaeology drawings.

Keywords: archaeology drawings, digital heritage, image generation, deep learning

Procedia PDF Downloads 33

Authors: Mebarek Boukelkoul, Abdelhalim Haroun

Abstract:

By means of the first principle calculation, we have investigated the structural, magnetic and magneto-optical properties of the ultra-thin films of Fen/Cu(001) with (n=1, 2, 3). We adopted a relativistic approach using DFT theorem with local spin density approximation (LSDA). The electronic structure is performed within the framework of the Spin-Polarized Relativistic (SPR) Linear Muffin-Tin Orbitals (LMTO) with the Atomic Sphere Approximation (ASA) method. During the variational principle, the crystal wave function is expressed as a linear combination of the Bloch sums of the so-called relativistic muffin-tin orbitals centered on the atomic sites. The crystalline structure is calculated after an atomic relaxation process using the optimization of the total energy with respect to the atomic interplane distance. A body-centered tetragonal (BCT) pseudomorphic crystalline structure with a tetragonality ratio c/a larger than unity is found. The magnetic behaviour is characterized by an enhanced magnetic moment and a ferromagnetic interplane coupling. The polar magneto-optical Kerr effect spectra are given over a photon energy range extended to 15eV and the microscopic origin of the most interesting features are interpreted by interband transitions. Unlike thin layers, the anisotropy in the ultra-thin films is characterized by a perpendicular magnetization which is perpendicular to the film plane.

Keywords: ultrathin films, magnetism, magneto-optics, pseudomorphic structure

Procedia PDF Downloads 314

Authors: Deborah Eric, Abbas Ahmad Khan

Abstract:

Lead halide perovskites quantum dots have attracted immense scientific and technological interest for successful photovoltaic applications because of their remarkable optoelectronic properties. In this paper, we have simulated CsMAPbI3-xBrx based quantum dots to implement their use in intermediate band solar cells (IBSC). These types of materials exhibit optical and electrical properties distinct from their bulk counterparts due to quantum confinement. The conceptual framework provides a route to analyze the electronic properties of quantum dots. This layer of quantum dots optimizes the position and bandwidth of IB that lies in the forbidden region of the conventional bandgap. A three-dimensional MAPbI3 quantum dot (QD) with geometries including spherical, cubic, and conical has been embedded in the CsPbBr3 matrix. Bound energy wavefunction gives rise to miniband, which results in the formation of IB. If there is more than one miniband, then there is a possibility of having more than one IB. The optimization of QD size results in more IBs in the forbidden region. One band time-independent Schrödinger equation using the effective mass approximation with step potential barrier is solved to compute the electronic states. Envelope function approximation with BenDaniel-Duke boundary condition is used in combination with the Schrödinger equation for the calculation of eigen energies and Eigen energies are solved for the quasi-bound states using an eigenvalue study. The transfer matrix method is used to study the quantum tunneling of MAPbI3 QD through neighbor barriers of CsPbI3. Electronic states are computed using Schrödinger equation with effective mass approximation by considering quantum dot and wetting layer assembly. Results have shown the varying the quantum dot size affects the energy pinning of QD. Changes in the ground, first, second state energies have been observed. The QD is non-zero at the center and decays exponentially to zero at boundaries. Quasi-bound states are characterized by envelope functions. It has been observed that conical quantum dots have maximum ground state energy at a small radius. Increasing the wetting layer thickness exhibits energy signatures similar to bulk material for each QD size.

Keywords: perovskite, intermediate bandgap, quantum dots, miniband formation

Procedia PDF Downloads 147

Authors: Bastian Vollrath, Hartwig Hubel

Abstract:

In case of over-elastic and cyclic loading, strain may accumulate due to a ratcheting mechanism until the state of shakedown is possibly achieved. Load history dependent numerical investigations by a step-by-step analysis are rather costly in terms of engineering time and numerical effort. In the case of multi-parameter loading, where various independent loadings affect the final state of shakedown, the computational effort becomes an additional challenge. Therefore, direct methods like the Simplified Theory of Plastic Zones (STPZ) are developed to solve the problem with a few linear elastic analyses. Post-shakedown quantities such as strain ranges and cyclic accumulated strains are calculated approximately by disregarding the load history. The STPZ is based on estimates of a transformed internal variable, which can be used to perform modified elastic analyses, where the elastic material parameters are modified, and initial strains are applied as modified loading, resulting in residual stresses and strains. The STPZ already turned out to work well with respect to cyclic loading between two states of loading. Usually, few linear elastic analyses are sufficient to obtain a good approximation to the post-shakedown quantities. In a multi-dimensional load domain, the approximation of the transformed internal variable transforms from a plane problem into a hyperspace problem, where time-consuming approximation methods need to be applied. Therefore, a solution restricted to structures with four stress components was developed to estimate the transformed internal variable by means of three-dimensional vector algebra. This paper presents the extension to cyclic multi-parameter loading so that an unlimited number of load cases can be taken into account. The theoretical basis and basic presumptions of the Simplified Theory of Plastic Zones are outlined for the case of elastic shakedown. The extension of the method to many load cases is explained, and a workflow of the procedure is illustrated. An example, adopting the FE-implementation of the method into ANSYS and considering multilinear hardening is given which highlights the advantages of the method compared to incremental, step-by-step analysis.

Keywords: cyclic loading, direct method, elastic shakedown, multi-parameter loading, STPZ

Procedia PDF Downloads 141

Authors: Belgherbi Aicha, Hadjidj Ismahen, Bessaid Abdelhafid

Abstract:

The phase of segmentation is an important step in the processing and interpretation of medical images. In this paper, we focus on the segmentation of liver and spleen from the abdomen computed tomography (CT) images. The importance of our study comes from the fact that the segmentation of ROI from CT images is usually a difficult task. This difficulty is the gray’s level of which is similar to the other organ also the ROI are connected to the ribs, heart, kidneys, etc. Our proposed method is based on the anatomical information and mathematical morphology tools used in the image processing field. At first, we try to remove the surrounding and connected organs and tissues by applying morphological filters. This first step makes the extraction of interest regions easier. The second step consists of improving the quality of the image gradient. In this step, we propose a method for improving the image gradient to reduce these deficiencies by applying the spatial filters followed by the morphological filters. Thereafter we proceed to the segmentation of the liver, spleen. To validate the segmentation technique proposed, we have tested it on several images. Our segmentation approach is evaluated by comparing our results with the manual segmentation performed by an expert. The experimental results are described in the last part of this work.The system has been evaluated by computing the sensitivity and specificity between the semi-automatically segmented (liver and spleen) contour and the manually contour traced by radiological experts.

Keywords: CT images, liver and spleen segmentation, anisotropic diffusion filter, morphological filters, watershed algorithm

Procedia PDF Downloads 470

Authors: Kausar Harun, Ahmad Azmin Mohamad

Abstract:

Zinc oxide (ZnO) has been extensively used in optoelectronic devices, with recent interest as photoanode material in dye-sensitize solar cell. Numerous methods employed to experimentally synthesized ZnO, while some are theoretically-modeled. Both approaches provide information on ZnO properties, but theoretical calculation proved to be more accurate and timely effective. Thus, integration between these two methods is essential to intimately resemble the properties of synthesized ZnO. In this study, experimentally-grown ZnO nanoparticles were prepared by sol-gel storage method with zinc acetate dihydrate and methanol as precursor and solvent. A 1 M sodium hydroxide (NaOH) solution was used as stabilizer. The optimum time to produce ZnO nanoparticles were recorded as 12 hours. Phase and structural analysis showed that single phase ZnO produced with wurtzite hexagonal structure. Further work on quantitative analysis was done via Rietveld-refinement method to obtain structural and crystallite parameter such as lattice dimensions, space group, and atomic coordination. The lattice dimensions were a=b=3.2498Å and c=5.2068Å which were later used as main input in first-principles calculations. By applying density-functional theory (DFT) embedded in CASTEP computer code, the structure of synthesized ZnO was built and optimized using several exchange-correlation functionals. The generalized-gradient approximation functional with Perdew-Burke-Ernzerhof and Hubbard U corrections (GGA-PBE+U) showed the structure with lowest energy and lattice deviations. In this study, emphasize also given to the modification of valence electron energy level to overcome the underestimation in DFT calculation. Both Zn and O valance energy were fixed at Ud=8.3 eV and Up=7.3 eV, respectively. Hence, the following electronic and optical properties of synthesized ZnO were calculated based on GGA-PBE+U functional within ultrasoft-pseudopotential method. In conclusion, the incorporation of Rietveld analysis into first-principles calculation was valid as the resulting properties were comparable with those reported in literature. The time taken to evaluate certain properties via physical testing was then eliminated as the simulation could be done through computational method.

Keywords: density functional theory, first-principles, Rietveld-refinement, ZnO nanoparticles

Procedia PDF Downloads 292

Authors: Simon Keller, Rainer Mueller

Abstract:

Continuous range queries are a common means to handle mobile clients in high-density areas. Most existing approaches focus on settings in which the range queries for location-based services are more or less static, whereas the mobile clients in the ranges move. We focus on a category called dynamic real-time range queries (DRRQ), assuming that both, clients requested by the query and the inquirers, are mobile. In consequence, the query parameters and the query results continuously change. This leads to two requirements: the ability to deal with an arbitrarily high number of mobile nodes (scalability) and the real-time delivery of range query results. In this paper, we present the highly decentralized solution adaptive quad streaming (AQS) for the requirements of DRRQs. AQS approximates the query results in favor of a controlled real-time delivery and guaranteed scalability. While prior works commonly optimize data structures on the involved servers, we use AQS to focus on a highly distributed cell structure without data structures automatically adapting to changing client distributions. Instead of the commonly used request-response approach, we apply a lightweight streaming method in which no bidirectional communication and no storage or maintenance of queries are required at all.

Keywords: approximation of client distributions, continuous spatial range queries, mobile objects, streaming-based decentralization in spatial mobile environments

Procedia PDF Downloads 120

Authors: Jack R. McKenzie, Peter A. Appleby, Thomas House, Neil Walton

Abstract:

Cold-start is a notoriously difficult problem which can occur in recommendation systems, and arises when there is insufficient information to draw inferences for users or items. To address this challenge, a contextual bandit algorithm – the Fast Approximate Bayesian Contextual Cold Start Learning algorithm (FAB-COST) – is proposed, which is designed to provide improved accuracy compared to the traditionally used Laplace approximation in the logistic contextual bandit, while controlling both algorithmic complexity and computational cost. To this end, FAB-COST uses a combination of two moment projection variational methods: Expectation Propagation (EP), which performs well at the cold start, but becomes slow as the amount of data increases; and Assumed Density Filtering (ADF), which has slower growth of computational cost with data size but requires more data to obtain an acceptable level of accuracy. By switching from EP to ADF when the dataset becomes large, it is able to exploit their complementary strengths. The empirical justification for FAB-COST is presented, and systematically compared to other approaches on simulated data. In a benchmark against the Laplace approximation on real data consisting of over 670, 000 impressions from autotrader.co.uk, FAB-COST demonstrates at one point increase of over 16% in user clicks. On the basis of these results, it is argued that FAB-COST is likely to be an attractive approach to cold-start recommendation systems in a variety of contexts.

Keywords: cold-start learning, expectation propagation, multi-armed bandits, Thompson Sampling, variational inference

Procedia PDF Downloads 95

Authors: Lifar M. S., Tereshchenko A. A., Bulgakov A. N., Guda S. A., Guda A. A., Soldatov A. V.

Abstract:

Metal nanoparticles are widely used as heterogeneous catalysts to activate adsorbed molecules and reduce the energy barrier of the reaction. Reaction product yield depends on the interplay between elementary processes - adsorption, activation, reaction, and desorption. These processes, in turn, depend on the inlet feed concentrations, temperature, and pressure. At stationary conditions, the active surface sites may be poisoned by reaction byproducts or blocked by thermodynamically adsorbed gaseous reagents. Thus, the yield of reaction products can significantly drop. On the contrary, the dynamic control accounts for the changes in the surface properties and adjusts reaction parameters accordingly. Therefore dynamic control may be more efficient than stationary control. In this work, a reinforcement learning algorithm has been applied to control the simulation of CO oxidation on a catalyst. The policy gradient algorithm is learned to maximize the CO₂ production rate based on the CO and O₂ flows at a given time step. Nonstationary solutions were found for the regime with surface deactivation. The maximal product yield was achieved for periodic variations of the gas flows, ensuring a balance between available adsorption sites and the concentration of activated intermediates. This methodology opens a perspective for the optimization of catalytic reactions under nonstationary conditions.

Keywords: artificial intelligence, catalyst, co oxidation, reinforcement learning, dynamic control

Procedia PDF Downloads 99

Authors: Bo-Bin Xiaoa, Chia-Yi Liua

Abstract:

In the task of searching for anomalous radiation sources, personnel holding radiation detectors to search for radiation sources may be exposed to unnecessary radiation risk, and automated search using machines becomes a required project. The research uses various sophisticated algorithms, which are double Q learning, dueling network, and NoisyNet, of deep reinforcement learning to search for radiation sources. The simulation environment, which is a 10*10 grid and one shielding wall setting in it, improves the development of the AI model by training 1 million episodes. In each episode of training, the radiation source position, the radiation source intensity, agent position, shielding wall position, and shielding wall length are all set randomly. The three algorithms are applied to run AI model training in four environments where the training shielding wall is a full-shielding wall, a lead wall, a concrete wall, and a lead wall or a concrete wall appearing randomly. The 12 best performance AI models are selected by observing the reward value during the training period and are evaluated by comparing these AI models with the gradient search algorithm. The results show that the performance of the AI model, no matter which one algorithm, is far better than the gradient search algorithm. In addition, the simulation environment becomes more complex, the AI model which applied Double DQN combined Dueling and NosiyNet algorithm performs better.

Keywords: double Q learning, dueling network, NoisyNet, source searching

Procedia PDF Downloads 88

Authors: Wanhyun Cho, Soonja Kang, Sangkyoon Kim, Soonyoung Park

Abstract:

In this paper, we propose a novel inference algorithm for the multi-class Gaussian process classification model that can be used in the field of human behavior recognition. This algorithm can drive simultaneously both a posterior distribution of a latent function and estimators of hyper-parameters in a Gaussian process classification model with multi-class. Our algorithm is based on the Laplace approximation (LA) technique and variational EM framework. This is performed in two steps: called expectation and maximization steps. First, in the expectation step, using the Bayesian formula and LA technique, we derive approximately the posterior distribution of the latent function indicating the possibility that each observation belongs to a certain class in the Gaussian process classification model. Second, in the maximization step, using a derived posterior distribution of latent function, we compute the maximum likelihood estimator for hyper-parameters of a covariance matrix necessary to define prior distribution for latent function. These two steps iteratively repeat until a convergence condition satisfies. Moreover, we apply the proposed algorithm with human action classification problem using a public database, namely, the KTH human action data set. Experimental results reveal that the proposed algorithm shows good performance on this data set.

Keywords: bayesian rule, gaussian process classification model with multiclass, gaussian process prior, human action classification, laplace approximation, variational EM algorithm

Procedia PDF Downloads 311

Authors: Syed M. Rizvi, Yiqing Meng, Simon Iwnicki

Abstract:

With an increased demand in the land cargo industry, it is predicted that the freight trade will rise to a record $1.1 trillion in revenue and volume in the following years to come. This increase is mainly driven by the e-commerce model ever so popular in the consumer market. Many innovative ideas have stemmed from this demand and change in lifestyle likes of which include e-bike cargo and drones. Rural and urban areas are facing air quality challenges to keep pollution levels in city centre to a minimum. For this purpose, this paper presents the design and implementation of a non-linear PID control system, employing a micro-controller and low cost sensing technique, for controlling an electrical motor based cargo vehicle with various loads, to follow a leading vehicle (bike). Within using this system, the cargo vehicle will have no load influence on the bike rider on different gradient conditions, such as hill climbing. The system is being integrated with a microcontroller to continuously measure several parameters such as relative displacement between bike and the cargo vehicle and gradient of the road, and process these measurements to create a portable controller capable of controlling the performance of electrical vehicle without the need of a PC. As a result, in the case of carrying 180kg of parcel weight, the cargo vehicle can maintain a reasonable spacing over a short length of sensor travel between the bike and itself.

Keywords: cargo, e-bike, microcontroller, embedded system, nonlinear pid, self-adaptive, inertial measurement unit (IMU)

Procedia PDF Downloads 189

Authors: Muhtasim Shafi Kader, Fizar Ahmed, Annesha Acharjee

Abstract:

Machine learning and data mining are crucial in health care, as well as medical information and detection. Machine learning approaches are now being utilized to improve awareness of a variety of critical health issues, including diabetes detection, neuron cell tumor diagnosis, COVID 19 identification, and so on. Parkinson’s disease is basically a disease for our senior citizens in Bangladesh. Parkinson's Disease indications often seem progressive and get worst with time. People got affected trouble walking and communicating with the condition advances. Patients can also have psychological and social vagaries, nap problems, hopelessness, reminiscence loss, and weariness. Parkinson's disease can happen in both men and women. Though men are affected by the illness at a proportion that is around partial of them are women. In this research, we have to get out the accurate ML algorithm to find out the disease with a predictable dataset and the model of the following machine learning classifiers. Therefore, nine ML classifiers are secondhand to portion study to use machine learning approaches like as follows, Naive Bayes, Adaptive Boosting, Bagging Classifier, Decision Tree Classifier, Random Forest classifier, XBG Classifier, K Nearest Neighbor Classifier, Support Vector Machine Classifier, and Gradient Boosting Classifier are used.

Keywords: naive bayes, adaptive boosting, bagging classifier, decision tree classifier, random forest classifier, XBG classifier, k nearest neighbor classifier, support vector classifier, gradient boosting classifier

Procedia PDF Downloads 112

Authors: Y. Wang, S. Zhang, X. Chen

Abstract:

The longitudinal compressive strength of fibre reinforced plastic (FRP) possess a large stochastic variability, which limits efficient application of composite structures. This study aims to address how the random fibre packing affects the uncertainty of FRP compressive strength. An novel approach is proposed to generate random fibre packing status by a combination of Latin hypercube sampling and random sequential expansion. 3D nonlinear finite element model is built which incorporates both the matrix plasticity and fibre geometrical instability. The matrix is modeled by isotropic ideal elasto-plastic solid elements, and the fibres are modeled by linear-elastic rebar elements. Composite with a series of different nominal fibre volume fractions are studied. Premature fibre waviness at different magnitude and direction is introduced in the finite element model. Compressive tests on uni-directional CFRP (carbon fibre reinforced plastic) are conducted following the ASTM D6641. By a comparison of 3D FE models and compressive tests, it is clearly shown that the stochastic variation of compressive strength is partly caused by the random fibre packing, and normal or lognormal distribution tends to be a good fit the probabilistic compressive strength. Furthermore, it is also observed that different random fibre packing could trigger two different fibre micro-buckling modes while subjected to longitudinal compression: out-of-plane buckling and twisted buckling. The out-of-plane buckling mode results much larger compressive strength, and this is the major reason why the random fibre packing results a large uncertainty in the FRP compressive strength. This study would contribute to new approaches to the quality control of FRP considering higher compressive strength or lower uncertainty.

Keywords: compressive strength, FRP, micro-buckling, random fibre packing

Procedia PDF Downloads 256

Authors: Jonathan Sohn

Abstract:

Deep neural networks (DNNs) have shown state-of-the-art performance for many applications, including computer vision, natural language processing, and speech recognition. Recently, adversarial attacks have been studied in the context of deep neural networks, which aim to alter the results of deep neural networks by modifying the inputs slightly. For example, an adversarial attack on a DNN used for object detection can cause the DNN to miss certain objects. As a result, the reliability of DNNs is undermined by their lack of robustness against adversarial attacks, raising concerns about their use in safety-critical applications such as autonomous driving. In this paper, we focus on studying the adversarial attacks and defenses on DNNs for image classification. There are two types of adversarial attacks studied which are fast gradient sign method (FGSM) attack and projected gradient descent (PGD) attack. A DNN forms decision boundaries that separate the input images into different categories. The adversarial attack slightly alters the image to move over the decision boundary, causing the DNN to misclassify the image. FGSM attack obtains the gradient with respect to the image and updates the image once based on the gradients to cross the decision boundary. PGD attack, instead of taking one big step, repeatedly modifies the input image with multiple small steps. There is also another type of attack called the target attack. This adversarial attack is designed to make the machine classify an image to a class chosen by the attacker. We can defend against adversarial attacks by incorporating adversarial examples in training. Specifically, instead of training the neural network with clean examples, we can explicitly let the neural network learn from the adversarial examples. In our experiments, the digit recognition accuracy on the MNIST dataset drops from 97.81% to 39.50% and 34.01% when the DNN is attacked by FGSM and PGD attacks, respectively. If we utilize FGSM training as a defense method, the classification accuracy greatly improves from 39.50% to 92.31% for FGSM attacks and from 34.01% to 75.63% for PGD attacks. To further improve the classification accuracy under adversarial attacks, we can also use a stronger PGD training method. PGD training improves the accuracy by 2.7% under FGSM attacks and 18.4% under PGD attacks over FGSM training. It is worth mentioning that both FGSM and PGD training do not affect the accuracy of clean images. In summary, we find that PGD attacks can greatly degrade the performance of DNNs, and PGD training is a very effective way to defend against such attacks. PGD attacks and defence are overall significantly more effective than FGSM methods.

Keywords: deep neural network, adversarial attack, adversarial defense, adversarial machine learning

Procedia PDF Downloads 173

Authors: Hossein Jafari, Xiangfang Li, Lijun Qian, Alexander Aved, Timothy Kroecker

Abstract:

Traditionally in sensor networks and recently in the Internet of Things, numerous heterogeneous sensors are deployed in distributed manner to monitor a phenomenon that often can be model by an underlying stochastic process. The big time-series data collected by the sensors must be analyzed to detect change in the stochastic process as quickly as possible with tolerable false alarm rate. However, sensors may have different accuracy and sensitivity range, and they decay along time. As a result, the big time-series data collected by the sensors will contain uncertainties and sometimes they are conflicting. In this study, we present a framework to take advantage of Evidence Theory (a.k.a. Dempster-Shafer and Dezert-Smarandache Theories) capabilities of representing and managing uncertainty and conflict to fast change detection and effectively deal with complementary hypotheses. Specifically, Kullback-Leibler divergence is used as the similarity metric to calculate the distances between the estimated current distribution with the pre- and post-change distributions. Then mass functions are calculated and related combination rules are applied to combine the mass values among all sensors. Furthermore, we applied the method to estimate the minimum number of sensors needed to combine, so computational efficiency could be improved. Cumulative sum test is then applied on the ratio of pignistic probability to detect and declare the change for decision making purpose. Simulation results using both synthetic data and real data from experimental setup demonstrate the effectiveness of the presented schemes.

Keywords: CUSUM, evidence theory, kl divergence, quickest change detection, time series data

Procedia PDF Downloads 310

Authors: Iacopo Carnacina, Mohammad Keshtpoor, Richard Yablonsky

Abstract:

Non-structural protection measures against flooding are becoming increasingly popular flood risk mitigation strategies. In particular, coastal flood insurance impacts not only private citizens but also insurance and reinsurance companies, who may require it to retain solvency and better understand the risks they face from a catastrophic coastal flood event. In this context, a framework is presented here to assess the risk for coastal flooding across the UK. The area has a long history of catastrophic flood events, including the Great Flood of 1953 and the 2013 Cyclone Xaver storm, both of which led to significant loss of life and property. The current framework will leverage a technology based on a hydrodynamic model (Delft3D Flexible Mesh). This flexible mesh technology, coupled with a calibration technique, allows for better utilisation of computational resources, leading to higher resolution and more detailed results. The generation of a stochastic set of extra tropical cyclone (ETC) events supports the evaluation of the financial losses for the whole area, also accounting for correlations between different locations in different scenarios. Finally, the solution shows a detailed analysis for the Thames River, leveraging the information available on flood barriers and levees. Two realistic disaster scenarios for the Greater London area are simulated: In the first scenario, the storm surge intensity is not high enough to fail London’s flood defences, but in the second scenario, London’s flood defences fail, highlighting the potential losses from a catastrophic coastal flood event.

Keywords: storm surge, stochastic model, levee failure, Thames River

Procedia PDF Downloads 218

Authors: Xuechen Chen

Abstract:

The EU-ASEAN relations which can be dated back to 1972 represents one of the oldest group-to-group relationship in international politics. Despite a longstanding dialogue partnership, the EU and ASEAN have long been reluctant to forge deeper and substantial cooperation in political and security domains. However, the year of 2012 witnessed a salient shift in EU-ASEAN relations, with the EU significantly elevating ASEAN's profile in its external relations. Given the limited scholarly attention that has been devoted to this change in ASEAN-EU relations, this article explores why there has been a greater level of engagement and approximation between the EU and ASEAN. In particular, it asks why the EU, which had long been reluctant to recognize ASEAN as a strategic partner, has changed its policy towards ASEAN. Drawing on social constructivism, this article argues that the EU’s and ASEAN’s evolving identity-formation processes have played a significant role in reshaping their mutual perceptions, which subsequently leads to the modification of the interregional policies of both actors. The methodology of this study is based on content analysis of a wide range of official documents and policy papers from the EU and ASEAN, as well as more than 20 in-depth elite interviews with diplomats and experts working on the EU-ASEAN relationship from both organisations. Departing from the existing works which mainly adopt a Eurocentric perspective when analysing the EU-ASEAN interregionalism, this study suggests that the approximation of the EU-ASEAN relationship between 2012 and 2017 is driven by both actors’ adjustment of international identities, together with the internal dynamics and systematic changes within both regions.

Keywords: Association of Southeast Asia Nations, European Union, EU foreign policy, interregionalism

Procedia PDF Downloads 133

Authors: Belgherbi Aicha, Hadjidj Ismahen, Bessaid Abdelhafid

Abstract:

Interpretation of medical images benefits from anatomical and physiological priors to optimize computer-aided diagnosis applications. Segmentation of liver, spleen and kidneys is regarded as a major primary step in the computer-aided diagnosis of abdominal organ diseases. In this paper, a semi-automated method for medical image data is presented for the abdominal organ segmentation data using mathematical morphology. Our proposed method is based on hierarchical segmentation and watershed algorithm. In our approach, a powerful technique has been designed to suppress over-segmentation based on mosaic image and on the computation of the watershed transform. Our algorithm is currency in two parts. In the first, we seek to improve the quality of the gradient-mosaic image. In this step, we propose a method for improving the gradient-mosaic image by applying the anisotropic diffusion filter followed by the morphological filters. Thereafter, we proceed to the hierarchical segmentation of the liver, spleen and kidney. To validate the segmentation technique proposed, we have tested it on several images. Our segmentation approach is evaluated by comparing our results with the manual segmentation performed by an expert. The experimental results are described in the last part of this work.

Keywords: anisotropic diffusion filter, CT images, morphological filter, mosaic image, simultaneous organ segmentation, the watershed algorithm

Procedia PDF Downloads 419

Authors: Alexandra Sargent, Sarah Ferris, Ioannis Theofanous

Abstract:

The Abbott RealTime High Risk HPV test (RealTime HPV) is one of five assays clinically validated and approved by the English NHS Cervical Screening Programme (CSP) for HPV triage of low grade dyskaryosis and test-of-cure of treated Cervical Intraepithelial Neoplasia. The assay is a highly automated multiplex real-time PCR test for detecting 14 high risk (hr) HPV types, with simultaneous differentiation of HPV 16 and HPV 18 versus non-HPV 16/18 hrHPV. An endogenous internal control ensures sample cellularity, controls extraction efficiency and PCR inhibition. The original cervical specimen collected in SurePath (SP) liquid-based cytology (LBC) medium (BD Diagnostics) and the SP post-gradient cell pellets (SPG) after cytological processing are both CE marked for testing with the RealTime HPV test. During the 2011 NHSCSP validation of new tests only the original aliquot of SP LBC medium was investigated. Residual sample volume left after cytology slide preparation is low and may not always have sufficient volume for repeat HPV testing or for testing of other biomarkers that may be implemented in testing algorithms in the future. The SPG samples, however, have sufficient volumes to carry out additional testing and necessary laboratory validation procedures. This study investigates the correlation of RealTime HPV results of cervical specimens collected in SP LBC medium from women with low grade cytological abnormalities observed with matched pairs of original SP LBC medium and SP post-gradient cell pellets (SPG) after cytology processing. Matched pairs of SP and SPG samples from 750 women with borderline (N = 392) and mild (N = 351) cytology were available for this study. Both specimen types were processed and parallel tested for the presence of hrHPV with RealTime HPV according to the manufacturer´s instructions. HrHPV detection rates and concordance between test results from matched SP and SPGCP pairs were calculated. A total of 743 matched pairs with valid test results on both sample types were available for analysis. An overall-agreement of hrHPV test results of 97.5% (k: 0.95) was found with matched SP/SPG pairs and slightly lower concordance (96.9%; k: 0.94) was observed on 392 pairs from women with borderline cytology compared to 351 pairs from women with mild cytology (98.0%; k: 0.95). Partial typing results were highly concordant in matched SP/SPG pairs for HPV 16 (99.1%), HPV 18 (99.7%) and non-HPV16/18 hrHPV (97.0%), respectively. 19 matched pairs were found with discrepant results: 9 from women with borderline cytology and 4 from women with mild cytology were negative on SPG and positive on SP; 3 from women with borderline cytology and 3 from women with mild cytology were negative on SP and positive on SPG. Excellent correlation of hrHPV DNA test results was found between matched pairs of SP original fluid and post-gradient cell pellets from women with low grade cytological abnormalities tested with the Abbott RealTime High-Risk HPV assay, demonstrating robust performance of the test with both specimen types and reassuring the utility of the assay for cytology triage with both specimen types.

Keywords: Abbott realtime test, HPV, SurePath liquid based cytology, surepath post-gradient cell pellet

Procedia PDF Downloads 231

Authors: Fatih Kılıç, Burak Birol, Necmettin Maraşlı

Abstract:

The directional solidification process is generally used for homogeneous compound production, single crystal growth, and refining (zone refining), etc. processes. The most important two parameters that control eutectic structures are temperature gradient and grain growth rate which are called as solidification parameters The solidification behavior and microstructure characteristics is an interesting topic due to their effects on the properties and performance of the alloys containing eutectic compositions. The solidification behavior of multicomponent and multiphase systems is an important parameter for determining various properties of these materials. The researches have been conducted mostly on the solidification of pure materials or alloys containing two phases. However, there are very few studies on the literature about multiphase reactions and microstructure formation of multicomponent alloys during solidification. Because of this situation, it is important to study the microstructure formation and the thermodynamical, thermophysical and microstructural properties of these alloys. The production process is difficult due to easy oxidation of magnesium and therefore, there is not a comprehensive study concerning alloys containing high Mg (> 30 wt.% Mg). With the increasing amount of Mg inside Al alloys, the specific weight decreases, and the strength shows a slight increase, while due to formation of β-Al8Mg5 phase, ductility lowers. For this reason, production, examination and development of high Mg containing alloys will initiate the production of new advanced engineering materials. The original value of this research can be described as obtaining high Mg containing (> 30% Mg) Al based multicomponent alloys by melting under vacuum; controlled directional solidification with various growth rates at a constant temperature gradient; and establishing relationship between solidification rate and microstructural, mechanical, electrical and thermal properties. Therefore, within the scope of this research, some > 30% Mg containing ternary or quaternary Al alloy compositions were determined, and it was planned to investigate the effects of directional solidification rate on the mechanical, electrical and thermal properties of these alloys. Within the scope of the research, the influence of the growth rate on microstructure parameters, microhardness, tensile strength, electrical conductivity and thermal conductivity of directionally solidified high Mg containing Al-32,2Mg-0,37Si; Al-30Mg-12Zn; Al-32Mg-1,7Ni; Al-32,2Mg-0,37Fe; Al-32Mg-1,7Ni-0,4Si; Al-33,3Mg-0,35Si-0,11Fe (wt.%) alloys with wide range of growth rate (50-2500 µm/s) and fixed temperature gradient, will be investigated. The work can be planned as; (a) directional solidification of Al-Mg based Al-Mg-Si, Al-Mg-Zn, Al-Mg-Ni, Al-Mg-Fe, Al-Mg-Ni-Si, Al-Mg-Si-Fe within wide range of growth rates (50-2500 µm/s) at a constant temperature gradient by Bridgman type solidification system, (b) analysis of microstructure parameters of directionally solidified alloys by using an optical light microscopy and Scanning Electron Microscopy (SEM), (c) measurement of microhardness and tensile strength of directionally solidified alloys, (d) measurement of electrical conductivity by four point probe technique at room temperature (e) measurement of thermal conductivity by linear heat flow method at room temperature.

Keywords: directional solidification, electrical conductivity, high Mg containing multicomponent Al alloys, microhardness, microstructure, tensile strength, thermal conductivity

Procedia PDF Downloads 243

Authors: Amr S. Fahil, Eman Ghoneim

Abstract:

Egypt made significant efforts during the past few years to discover significant renewable energy sources. Regions in Egypt that have been identified for geothermal potential investigation include the Gulf of Suez and the Western Desert. One of the most promising sites for the development of Egypt's Northern Western Desert is Siwa Oasis. The geological setting of the oasis, a tectonically generated depression situated in the northernmost region of the Western desert, supports the potential for substantial geothermal resources. Field data obtained from 27 deep oil wells along the Western Desert included bottom-hole temperature (BHT) depth to basement measurements, and geological maps; data were utilized in this study. The major lithological units, elevation, surface gradient, lineaments density, and remote sensing multispectral and topographic were mapped together to generate the related physiographic variables. Eleven thematic layers were integrated in a geographic information system (GIS) to create geothermal maps to aid in the detection of significant potential geothermal spots along the Siwa Oasis and its vicinity. The contribution of total magnetic intensity data with reduction to the pole (RTP) to the first investigation of the geothermal potential in Siwa Oasis is applied in this work. The integration of geospatial data with magnetic field measurements showed a clear correlation between areas of high heat flow and magnetic anomalies. Such anomalies can be interpreted as related to the existence of high geothermal energy and dense rock, which also have high magnetic susceptibility. The outcomes indicated that the study area has a geothermal gradient ranging from 18 to 42 °C/km, a heat flow ranging from 24.7 to 111.3 m.W. k−1, a thermal conductivity of 1.3–2.65 W.m−1.k−1 and a measured amplitude temperature maximum of 100.7 °C. The southeastern part of the Siwa Oasis, and some sporadic locations on the eastern section of the oasis were found to have significant geothermal potential; consequently, this location is suitable for future geothermal investigation. The adopted method might be applied to identify significant prospective geothermal energy locations in other regions of Egypt and East Africa.

Keywords: magnetic data, SRTM, depth to basement, remote sensing, GIS, geothermal gradient, heat flow, thermal conductivity

Procedia PDF Downloads 85

Authors: Simone Leonardi, Luca Ardito

Abstract:

Micro-influencers are effective elements in the marketing strategies of companies and institutions because of their capability to create an hyper-engaged audience around a specific topic of interest. In recent years, many scientific approaches and commercial tools have handled the task of detecting this type of social media users. These strategies adopt solutions ranging from rule based machine learning models to deep neural networks and graph analysis on text, images, and account information. This work compares the existing solutions and proposes an ensemble method to generalize them with different input data and social media platforms. The deployed solution combines deep learning models on unstructured data with statistical machine learning models on structured data. We retrieve both social media accounts information and multimedia posts on Twitter and Instagram. These data are mapped into feature vectors for an eXtreme Gradient Boosting (XGBoost) classifier. Sixty different topics have been analyzed to build a rule based gold standard dataset and to compare the performances of our approach against baseline classifiers. We prove the effectiveness of our work by comparing the accuracy, precision, recall, and f1 score of our model with different configurations and architectures. We obtained an accuracy of 0.91 with our best performing model.

Keywords: deep learning, gradient boosting, image processing, micro-influencers, NLP, social media

Procedia PDF Downloads 154

Authors: L. I. Izhar, T. Stathaki, K. Howell

Abstract:

Analyzing and interpreting of thermograms have been increasingly employed in the diagnosis and monitoring of diseases thanks to its non-invasive, non-harmful nature and low cost. In this paper, a novel system is proposed to improve diagnosis and monitoring of morphoea skin disorder based on integration with the published lines of Blaschko. In the proposed system, image registration based on global and local registration methods are found inevitable. This paper presents a modified normalized gradient cross-correlation (NGC) method to reduce large geometrical differences between two multimodal images that are represented by smooth gray edge maps is proposed for the global registration approach. This method is improved further by incorporating an iterative-based normalized cross-correlation coefficient (NCC) method. It is found that by replacing the final registration part of the NGC method where translational differences are solved in the spatial Fourier domain with the NCC method performed in the spatial domain, the performance and robustness of the NGC method can be greatly improved. It is shown in this paper that the hybrid NGC method not only outperforms phase correlation (PC) method but also improved misregistration due to translation, suffered by the modified NGC method alone for thermograms with ill-defined jawline. This also demonstrates that by using the gradients of the gray edge maps and a hybrid technique, the performance of the PC based image registration method can be greatly improved.

Keywords: Blaschko’s lines, image registration, morphoea, thermal imaging

Procedia PDF Downloads 291

Authors: Ana Clara Santos, Maria Manuela Portela, Bettina Schaefli

Abstract:

This work assesses the performance of an analytical model framework to generate daily flow duration curves, FDCs, based on climatic characteristics of the catchments and on their streamflow recession coefficients. According to the analytical model framework, precipitation is considered to be a stochastic process, modeled as a marked Poisson process, and recession is considered to be deterministic, with parameters that can be computed based on different models. The analytical model framework was tested for three case studies with different hydrological regimes located in Switzerland: pluvial, snow-dominated and glacier. For that purpose, five time intervals were analyzed (the four meteorological seasons and the civil year) and two developments of the model were tested: one considering a linear recession model and the other adopting a nonlinear recession model. Those developments were combined with recession coefficients obtained from two different approaches: forward and inverse estimation. The performance of the analytical framework when considering forward parameter estimation is poor in comparison with the inverse estimation for both, linear and nonlinear models. For the pluvial catchment, the inverse estimation shows exceptional good results, especially for the nonlinear model, clearing suggesting that the model has the ability to describe FDCs. For the snow-dominated and glacier catchments the seasonal results are better than the annual ones suggesting that the model can describe streamflows in those conditions and that future efforts should focus on improving and combining seasonal curves instead of considering single annual ones.

Keywords: analytical streamflow distribution, stochastic process, linear and non-linear recession, hydrological modelling, daily discharges

Procedia PDF Downloads 139

Authors: M. Mulumpwa, W. W. L. Jere, M. Lazaro, A. H. N. Mtethiwa

Abstract:

The study considered Seasonal Autoregressive Integrated Moving Average (SARIMA) processes to select an appropriate stochastic model to forecast the monthly data from the Lake Malawi water levels for the period 1986 through 2015. The appropriate model was chosen based on SARIMA (p, d, q) (P, D, Q)S. The Autocorrelation function (ACF), Partial autocorrelation (PACF), Akaike Information Criteria (AIC), Bayesian Information Criterion (BIC), Box–Ljung statistics, correlogram and distribution of residual errors were estimated. The SARIMA (1, 1, 0) (1, 1, 1)12 was selected to forecast the monthly data of the Lake Malawi water levels from August, 2015 to December, 2021. The plotted time series showed that the Lake Malawi water levels are decreasing since 2010 to date but not as much as was the case in 1995 through 1997. The future forecast of the Lake Malawi water levels until 2021 showed a mean of 474.47 m ranging from 473.93 to 475.02 meters with a confidence interval of 80% and 90% against registered mean of 473.398 m in 1997 and 475.475 m in 1989 which was the lowest and highest water levels in the lake respectively since 1986. The forecast also showed that the water levels of Lake Malawi will drop by 0.57 meters as compared to the mean water levels recorded in the previous years. These results suggest that the Lake Malawi water level may not likely go lower than that recorded in 1997. Therefore, utilisation and management of water-related activities and programs among others on the lake should provide room for such scenarios. The findings suggest a need to manage the Lake Malawi jointly and prudently with other stakeholders starting from the catchment area. This will reduce impacts of anthropogenic activities on the lake’s water quality, water level, aquatic and adjacent terrestrial ecosystems thereby ensuring its resilience to climate change impacts.

Keywords: forecasting, Lake Malawi, water levels, water level fluctuation, climate change, anthropogenic activities

Procedia PDF Downloads 204

Authors: Ajith Kumar S., Sankaran Namboothiri, Sankrish J., SarathKumar S., S. Anil Lal

Abstract:

A numerical analysis of flow over a heated circular cylinder is done in this paper. The governing equations, Navier-Stokes, and energy equation within the Boussinesq approximation along with continuity equation are solved using hybrid FEM-FVM technique. The density gradient created due to the heating of the cylinder will induce buoyancy force, opposite to the direction of action of acceleration due to gravity, g. In the present work, the flow direction and the direction of buoyancy force are taken as same (vertical flow configuration), so that the buoyancy force accelerates the mean flow past the cylinder. The relative dominance of the buoyancy force over the inertia force is characterized by the Richardson number (Ri), which is one of the parameter that governs the flow dynamics and heat transfer in this analysis. It is well known that above a certain value of Reynolds number, Re (ratio of inertia force over the viscous forces), the unsteady Von Karman vortices can be seen shedding behind the cylinder. The shedding wake patterns could be seriously altered by heating/cooling the cylinder. The non-dimensional shedding frequency called the Strouhal number is found to be increasing as Ri increases. The aerodynamic force coefficients CL and CD are observed to change its value. In the present vertical configuration of flow over the cylinder, as Ri increases, shedding frequency gets increased and suddenly drops down to zero at a critical value of Richardson number. The unsteady vortices turn to steady standing recirculation bubbles behind the cylinder after this critical Richardson number. This phenomenon is well known in literature as "Breakdown of the Karman Vortex Street". It is interesting to see the flow structures on further increase in the Richardson number. On further heating of the cylinder surface, the size of the recirculation bubble decreases without loosing its symmetry about the horizontal axis passing through the center of the cylinder. The separation angle is found to be decreasing with Ri. Finally, we observed a second critical Richardson number, after which the the flow will be attached to the cylinder surface without any wake behind it. The flow structures will be symmetrical not only about the horizontal axis, but also with the vertical axis passing through the center of the cylinder. At this stage, there will be a "single plume" emanating from the rear stagnation point of the cylinder. We also observed the transition of the plume is a strong function of the Richardson number.

Keywords: drag reduction, flow over circular cylinder, flow control, mixed convection flow, vortex shedding, vortex breakdown

Procedia PDF Downloads 382

Authors: S. Rozhkova, O. Rozhkova, A. Harlova, V. Lasukov

Abstract:

For optimal unbiased filter as mean-square and in the case of functioning anomalous noises in the observation memory channel, we have proved insensitivity of filter to inaccurate knowledge of the anomalous noise intensity matrix and its equivalence to truncated filter plotted only by non anomalous components of an observation vector.

Keywords: mathematical expectation, filtration, anomalous noise, memory

Procedia PDF Downloads 342