Search results for: Dirichlet kernel

Authors: Aprajeeta Jha, Punyadarshini P. Tripathy

Abstract:

Drying technologies for various food processing operations shares an inevitable linkage with energy, cost and environmental sustainability. Hence, solar drying of food grains has become imperative choice to combat duo challenges of meeting high energy demand for drying and to address climate change scenario. But performance and reliability of solar dryers depend hugely on sunshine period, climatic conditions, therefore, offer a limited control over drying conditions and have lower efficiencies. Solar drying technology, supported by Photovoltaic (PV) power plant and hybrid type solar air collector can potentially overpower the disadvantages of solar dryers. For development of such robust hybrid dryers; to ensure quality and shelf-life of paddy grains the optimization of process parameter becomes extremely critical. Investigation of the moisture distribution profile within the grains becomes necessary in order to avoid over drying or under drying of food grains in hybrid solar dryer. Computational simulations based on finite element modeling can serve as potential tool in providing a better insight of moisture migration during drying process. Hence, present work aims at optimizing the process parameters and to develop a 3-dimensional (3D) finite element model (FEM) for predicting moisture profile in paddy during solar drying. COMSOL Multiphysics was employed to develop a 3D finite element model for predicting moisture profile. Furthermore, optimization of process parameters (power level, air velocity and moisture content) was done using response surface methodology in design expert software. 3D finite element model (FEM) for predicting moisture migration in single kernel for every time step has been developed and validated with experimental data. The mean absolute error (MAE), mean relative error (MRE) and standard error (SE) were found to be 0.003, 0.0531 and 0.0007, respectively, indicating close agreement of model with experimental results. Furthermore, optimized process parameters for drying paddy were found to be 700 W, 2.75 m/s at 13% (wb) with optimum temperature, milling yield and drying time of 42˚C, 62%, 86 min respectively, having desirability of 0.905. Above optimized conditions can be successfully used to dry paddy in PV integrated solar dryer in order to attain maximum uniformity, quality and yield of product. PV-integrated hybrid solar dryers can be employed as potential and cutting edge drying technology alternative for sustainable energy and food security.

Keywords: finite element modeling, moisture migration, paddy grain, process optimization, PV integrated hybrid solar dryer

Procedia PDF Downloads 117

Authors: Elia Favarelli, Enrico Testi, Andrea Giorgetti

Abstract:

Autonomous structural health monitoring (SHM) of many structures and bridges became a topic of paramount importance for maintenance purposes and safety reasons. This paper proposes a set of machine learning (ML) tools to perform automatic feature selection and detection of anomalies in a bridge from vibrational data and compare different feature extraction schemes to increase the accuracy and reduce the amount of data collected. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric data in both standard and damaged conditions. The proposed framework starts from the first four fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by density-based time-domain filtering (tracking). The fundamental frequencies extracted are then fed to a dimensionality reduction block implemented through two different approaches: feature selection (intelligent multiplexer) that tries to estimate the most reliable frequencies based on the evaluation of some statistical features (i.e., mean value, variance, kurtosis), and feature extraction (auto-associative neural network (ANN)) that combine the fundamental frequencies to extract new damage sensitive features in a low dimensional feature space. Finally, one class classifier (OCC) algorithms perform anomaly detection, trained with standard condition points, and tested with normal and anomaly ones. In particular, a new anomaly detector strategy is proposed, namely one class classifier neural network two (OCCNN2), which exploit the classification capability of standard classifiers in an anomaly detection problem, finding the standard class (the boundary of the features space in normal operating conditions) through a two-step approach: coarse and fine boundary estimation. The coarse estimation uses classics OCC techniques, while the fine estimation is performed through a feedforward neural network (NN) trained that exploits the boundaries estimated in the coarse step. The detection algorithms vare then compared with known methods based on principal component analysis (PCA), kernel principal component analysis (KPCA), and auto-associative neural network (ANN). In many cases, the proposed solution increases the performance with respect to the standard OCC algorithms in terms of F1 score and accuracy. In particular, by evaluating the correct features, the anomaly can be detected with accuracy and an F1 score greater than 96% with the proposed method.

Keywords: anomaly detection, frequencies selection, modal analysis, neural network, sensor network, structural health monitoring, vibration measurement

Procedia PDF Downloads 90

Authors: Jaeyoung Lee

Abstract:

Autonomous driving systems require high reliability to provide people with a safe and comfortable driving experience. However, despite the development of a number of vehicle sensors, it is difficult to always provide high perceived performance in driving environments that vary from time to season. The image segmentation method using deep learning, which has recently evolved rapidly, provides high recognition performance in various road environments stably. However, since the system controls a vehicle in real time, a highly complex deep learning network cannot be used due to time and memory constraints. Moreover, efficient networks are optimized for GPU environments, which degrade performance in embedded processor environments equipped simple hardware accelerators. In this paper, a semantic segmentation network, matrix multiplication accelerator network (MMANet), optimized for matrix multiplication accelerator (MMA) on Texas instrument digital signal processors (TI DSP) is proposed to improve the recognition performance of autonomous driving system. The proposed method is designed to maximize the number of layers that can be performed in a limited time to provide reliable driving environment information in real time. First, the number of channels in the activation map is fixed to fit the structure of MMA. By increasing the number of parallel branches, the lack of information caused by fixing the number of channels is resolved. Second, an efficient convolution is selected depending on the size of the activation. Since MMA is a fixed, it may be more efficient for normal convolution than depthwise separable convolution depending on memory access overhead. Thus, a convolution type is decided according to output stride to increase network depth. In addition, memory access time is minimized by processing operations only in L3 cache. Lastly, reliable contexts are extracted using the extended atrous spatial pyramid pooling (ASPP). The suggested method gets stable features from an extended path by increasing the kernel size and accessing consecutive data. In addition, it consists of two ASPPs to obtain high quality contexts using the restored shape without global average pooling paths since the layer uses MMA as a simple adder. To verify the proposed method, an experiment is conducted using perfsim, a timing simulator, and the Cityscapes validation sets. The proposed network can process an image with 640 x 480 resolution for 6.67 ms, so six cameras can be used to identify the surroundings of the vehicle as 20 frame per second (FPS). In addition, it achieves 73.1% mean intersection over union (mIoU) which is the highest recognition rate among embedded networks on the Cityscapes validation set.

Keywords: edge network, embedded network, MMA, matrix multiplication accelerator, semantic segmentation network

Procedia PDF Downloads 95

Authors: Soyon Kim, Edward Kim

Abstract:

In order to retrieve information from the massive stream of songs in the music industry, music search by title, lyrics, artist, mood, and genre has become more important. Despite the subjectivity and controversy over the definition of music genres across different nations and cultures, automatic genre classification systems that facilitate the process of music categorization have been developed. Manual genre selection by music producers is being provided as statistical data for designing automatic genre classification systems. In this paper, an automatic music genre classification system utilizing non-negative matrix factorization (NMF) is proposed. Short-term characteristics of the music signal can be captured based on the timbre features such as mel-frequency cepstral coefficient (MFCC), decorrelated filter bank (DFB), octave-based spectral contrast (OSC), and octave band sum (OBS). Long-term time-varying characteristics of the music signal can be summarized with (1) the statistical features such as mean, variance, minimum, and maximum of the timbre features and (2) the modulation spectrum features such as spectral flatness measure, spectral crest measure, spectral peak, spectral valley, and spectral contrast of the timbre features. Not only these conventional basic long-term feature vectors, but also NMF based feature vectors are proposed to be used together for genre classification. In the training stage, NMF basis vectors were extracted for each genre class. The NMF features were calculated in the log spectral magnitude domain (NMF-LSM) as well as in the basic feature vector domain (NMF-BFV). For NMF-LSM, an entire full band spectrum was used. However, for NMF-BFV, only low band spectrum was used since high frequency modulation spectrum of the basic feature vectors did not contain important information for genre classification. In the test stage, using the set of pre-trained NMF basis vectors, the genre classification system extracted the NMF weighting values of each genre as the NMF feature vectors. A support vector machine (SVM) was used as a classifier. The GTZAN multi-genre music database was used for training and testing. It is composed of 10 genres and 100 songs for each genre. To increase the reliability of the experiments, 10-fold cross validation was used. For a given input song, an extracted NMF-LSM feature vector was composed of 10 weighting values that corresponded to the classification probabilities for 10 genres. An NMF-BFV feature vector also had a dimensionality of 10. Combined with the basic long-term features such as statistical features and modulation spectrum features, the NMF features provided the increased accuracy with a slight increase in feature dimensionality. The conventional basic features by themselves yielded 84.0% accuracy, but the basic features with NMF-LSM and NMF-BFV provided 85.1% and 84.2% accuracy, respectively. The basic features required dimensionality of 460, but NMF-LSM and NMF-BFV required dimensionalities of 10 and 10, respectively. Combining the basic features, NMF-LSM and NMF-BFV together with the SVM with a radial basis function (RBF) kernel produced the significantly higher classification accuracy of 88.3% with a feature dimensionality of 480.

Keywords: mel-frequency cepstral coefficient (MFCC), music genre classification, non-negative matrix factorization (NMF), support vector machine (SVM)

Procedia PDF Downloads 264

Authors: A. Chatel, I. S. Torreguitart, T. Verstraete

Abstract:

The paper deals with a single point optimization of the LS89 turbine using an adjoint optimization and defining the design variables within a CAD system. The advantage of including the CAD model in the design system is that higher level constraints can be imposed on the shape, allowing the optimized model or component to be manufactured. However, CAD-based approaches restrict the design space compared to node-based approaches where every node is free to move. In order to preserve a rich design space, we develop a methodology to refine the CAD model during the optimization and to create the best parameterization to use at each time. This study presents a methodology to progressively refine the design space, which combines parametric effectiveness with a differential evolutionary algorithm in order to create an optimal parameterization. In this manuscript, we show that by doing the parameterization at the CAD level, we can impose higher level constraints on the shape, such as the axial chord length, the trailing edge radius and G2 geometric continuity between the suction side and pressure side at the leading edge. Additionally, the adjoint sensitivities are filtered out and only smooth shapes are produced during the optimization process. The use of algorithmic differentiation for the CAD kernel and grid generator allows computing the grid sensitivities to machine accuracy and avoid the limited arithmetic precision and the truncation error of finite differences. Then, the parametric effectiveness is computed to rate the ability of a set of CAD design parameters to produce the design shape change dictated by the adjoint sensitivities. During the optimization process, the design space is progressively enlarged using the knot insertion algorithm which allows introducing new control points whilst preserving the initial shape. The position of the inserted knots is generally assumed. However, this assumption can hinder the creation of better parameterizations that would allow producing more localized shape changes where the adjoint sensitivities dictate. To address this, we propose using a differential evolutionary algorithm to maximize the parametric effectiveness by optimizing the location of the inserted knots. This allows the optimizer to gradually explore larger design spaces and to use an optimal CAD-based parameterization during the course of the optimization. The method is tested on the LS89 turbine cascade and large aerodynamic improvements in the entropy generation are achieved whilst keeping the exit flow angle fixed. The trailing edge and axial chord length, which are kept fixed as manufacturing constraints. The optimization results show that the multilevel optimizations were more efficient than the single level optimization, even though they used the same number of design variables at the end of the multilevel optimizations. Furthermore, the multilevel optimization where the parameterization is created using the optimal knot positions results in a more efficient strategy to reach a better optimum than the multilevel optimization where the position of the knots is arbitrarily assumed.

Keywords: adjoint, CAD, knots, multilevel, optimization, parametric effectiveness

Procedia PDF Downloads 86

Authors: Dmitriy Babenko, Sergey Yegorov, Ilya Korshukov, Aidana Sultanbekova, Valentina Barkhanskaya, Tatiana Bashirova, Yerzhan Zhunusov, Yevgeniya Li, Viktoriya Parakhina, Svetlana Kolesnichenko, Yeldar Baiken, Aruzhan Pralieva, Zhibek Zhumadilova, Matthew S. Miller, Gonzalo H. Hortelano, Anar Turmuhambetova, Antonella E. Chesca, Irina Kadyrova

Abstract:

Introduction: The rapidly evolving COVID-19 pandemic, along with the re-emergence of pathogens causing acute respiratory infections (ARI), has necessitated the development of novel diagnostic tools to differentiate various causes of ARI. MALDI-MS, due to its wide usage and affordability, has been proposed as a potential instrument for diagnosing SARS-CoV-2 versus non-SARS-CoV-2 ARI. The aim of this study was to investigate the potential of MALDI-MS in conjunction with a machine learning model to accurately distinguish between symptomatic infections caused by SARS-CoV-2 and non-SARS-CoV-2 during both the early and later phases of the pandemic. Furthermore, this study aimed to analyze mass spectrometry (MS) data obtained from nasal swabs of healthy individuals. Methods: We gathered mass spectra from 252 samples, comprising 108 SARS-CoV-2-positive samples obtained in 2020 (Covid 2020), 7 SARS-CoV- 2-positive samples obtained in 2023 (Covid 2023), 71 samples from symptomatic individuals without SARS-CoV-2 (Control non-Covid ARVI), and 66 samples from healthy individuals (Control healthy). All the samples were subjected to RT-PCR testing. For data analysis, we employed the caret R package to train and test seven machine-learning algorithms: C5.0, KNN, NB, RF, SVM-L, SVM-R, and XGBoost. We conducted a training process using a five-fold (outer) nested repeated (five times) ten-fold (inner) cross-validation with a randomized stratified splitting approach. Results: In this study, we utilized the Covid 2020 dataset as a case group and the non-Covid ARVI dataset as a control group to train and test various machine learning (ML) models. Among these models, XGBoost and SVM-R demonstrated the highest performance, with accuracy values of 0.97 [0.93, 0.97] and 0.95 [0.95; 0.97], specificity values of 0.86 [0.71; 0.93] and 0.86 [0.79; 0.87], and sensitivity values of 0.984 [0.984; 1.000] and 1.000 [0.968; 1.000], respectively. When examining the Covid 2023 dataset, the Naive Bayes model achieved the highest classification accuracy of 43%, while XGBoost and SVM-R achieved accuracies of 14%. For the healthy control dataset, the accuracy of the models ranged from 0.27 [0.24; 0.32] for k-nearest neighbors to 0.44 [0.41; 0.45] for the Support Vector Machine with a radial basis function kernel. Conclusion: Therefore, ML models trained on MALDI MS of nasopharyngeal swabs obtained from patients with Covid during the initial phase of the pandemic, as well as symptomatic non-Covid individuals, showed excellent classification performance, which aligns with the results of previous studies. However, when applied to swabs from healthy individuals and a limited sample of patients with Covid in the late phase of the pandemic, ML models exhibited lower classification accuracy.

Keywords: SARS-CoV-2, MALDI-TOF MS, ML models, nasopharyngeal swabs, classification

Procedia PDF Downloads 71

Authors: Merve Begum Terzi, Orhan Arikan, Adnan Abaci, Mustafa Candemir

Abstract:

Acute myocardial infarction is a major cause of death in the world. Therefore, its fast and reliable diagnosis is a major clinical need. ECG is the most important diagnostic methodology which is used to make decisions about the management of the cardiovascular diseases. In patients with acute myocardial ischemia, temporary chest pains together with changes in ST segment and T wave of ECG occur shortly before the start of myocardial infarction. In this study, a technique which detects changes in ST/T sections of ECG is developed for the early diagnosis of acute myocardial ischemia. For this purpose, a database of real ECG recordings that contains a set of records from 75 patients presenting symptoms of chest pain who underwent elective percutaneous coronary intervention (PCI) is constituted. 12-lead ECG’s of the patients were recorded before and during the PCI procedure. Two ECG epochs, which are the pre-inflation ECG which is acquired before any catheter insertion and the occlusion ECG which is acquired during balloon inflation, are analyzed for each patient. By using pre-inflation and occlusion recordings, ECG features that are critical in the detection of acute myocardial ischemia are identified and the most discriminative features for the detection of acute myocardial ischemia are extracted. A classification technique based on support vector machine (SVM) approach operating with linear and radial basis function (RBF) kernels to detect ischemic events by using ST-T derived joint features from non-ischemic and ischemic states of the patients is developed. The dataset is randomly divided into training and testing sets and the training set is used to optimize SVM hyperparameters by using grid-search method and 10fold cross-validation. SVMs are designed specifically for each patient by tuning the kernel parameters in order to obtain the optimal classification performance results. As a result of implementing the developed classification technique to real ECG recordings, it is shown that the proposed technique provides highly reliable detections of the anomalies in ECG signals. Furthermore, to develop a detection technique that can be used in the absence of ECG recording obtained during healthy stage, the detection of acute myocardial ischemia based on ECG recordings of the patients obtained during ischemia is also investigated. For this purpose, a Gaussian mixture model (GMM) is used to represent the joint pdf of the most discriminating ECG features of myocardial ischemia. Then, a Neyman-Pearson type of approach is developed to provide detection of outliers that would correspond to acute myocardial ischemia. Neyman – Pearson decision strategy is used by computing the average log likelihood values of ECG segments and comparing them with a range of different threshold values. For different discrimination threshold values and number of ECG segments, probability of detection and probability of false alarm values are computed, and the corresponding ROC curves are obtained. The results indicate that increasing number of ECG segments provide higher performance for GMM based classification. Moreover, the comparison between the performances of SVM and GMM based classification showed that SVM provides higher classification performance results over ECG recordings of considerable number of patients.

Keywords: ECG classification, Gaussian mixture model, Neyman–Pearson approach, support vector machine

Procedia PDF Downloads 118

Authors: Ja-Keoung Koo, Kensuke Nakamura, Hyohun Kim, Dongwha Shin, Yeonseok Kim, Ji-Su Ahn, Byung-Woo Hong

Abstract:

The machine learning techniques based on a convolutional neural network (CNN) have been actively developed and successfully applied to a variety of image analysis tasks including reconstruction, noise reduction, resolution enhancement, segmentation, motion estimation, object recognition. The classical visual information processing that ranges from low level tasks to high level ones has been widely developed in the deep learning framework. It is generally considered as a challenging problem to derive visual interpretation from high dimensional imagery data. A CNN is a class of feed-forward artificial neural network that usually consists of deep layers the connections of which are established by a series of non-linear operations. The CNN architecture is known to be shift invariant due to its shared weights and translation invariance characteristics. However, it is often computationally intractable to optimize the network in particular with a large number of convolution layers due to a large number of unknowns to be optimized with respect to the training set that is generally required to be large enough to effectively generalize the model under consideration. It is also necessary to limit the size of convolution kernels due to the computational expense despite of the recent development of effective parallel processing machinery, which leads to the use of the constantly small size of the convolution kernels throughout the deep CNN architecture. However, it is often desired to consider different scales in the analysis of visual features at different layers in the network. Thus, we propose a CNN model where different sizes of the convolution kernels are applied at each layer based on the random projection. We apply random filters with varying sizes and associate the filter responses with scalar weights that correspond to the standard deviation of the random filters. We are allowed to use large number of random filters with the cost of one scalar unknown for each filter. The computational cost in the back-propagation procedure does not increase with the larger size of the filters even though the additional computational cost is required in the computation of convolution in the feed-forward procedure. The use of random kernels with varying sizes allows to effectively analyze image features at multiple scales leading to a better generalization. The robustness and effectiveness of the proposed CNN based on random kernels are demonstrated by numerical experiments where the quantitative comparison of the well-known CNN architectures and our models that simply replace the convolution kernels with the random filters is performed. The experimental results indicate that our model achieves better performance with less number of unknown weights. The proposed algorithm has a high potential in the application of a variety of visual tasks based on the CNN framework. Acknowledgement—This work was supported by the MISP (Ministry of Science and ICT), Korea, under the National Program for Excellence in SW (20170001000011001) supervised by IITP, and NRF-2014R1A2A1A11051941, NRF2017R1A2B4006023.

Keywords: deep learning, convolutional neural network, random kernel, random projection, dimensionality reduction, object recognition

Procedia PDF Downloads 258

Authors: Noor S. Nasri, Eric C. A. Tatt, Usman D. Hamza, Jibril Mohammed, Husna M. Zain

Abstract:

Climate change has becoming a global environmental issue that may trigger irreversible changes in the environment with catastrophic consequences for human, animals and plants on our planet. Methane, carbon dioxide and nitrous oxide are the greenhouse gases (GHG) and as the main factor that significantly contributes to the global warming. Mainly carbon dioxide be produced and released to atmosphere by thermal industrial and power generation sectors. Methane is dominant component of natural gas releases significant of thermal heat, and the gaseous pollutants when homogeneous thermal combustion takes place at high temperature. Heterogeneous catalytic Combustion (HCC) principle is promising technologies towards environmental friendly energy production should be developed to ensure higher yields with lower pollutants gaseous emissions and perform complete combustion oxidation at moderate temperature condition as comparing to homogeneous high thermal combustion. Hence the principle has become a very interesting alternative total oxidation for the treatment of pollutants gaseous emission especially NOX product formation. Noble metals are dispersed on a support-porous HCC such as γ- Al2O3, TiO2 and ThO2 to increase thermal stability of catalyst and to increase to effectiveness of catalytic combustion. Support-porous HCC material to be selected based on factors of the surface area, porosity, thermal stability, thermal conductivity, reactivity with reactants or products, chemical stability, catalytic activity, and catalyst life. γ- Al2O3 with high catalytic activity and can last longer life of catalyst, is commonly used as the support for Pd catalyst at low temperatures. Sustainable and renewable support-material of bio-mass char was derived from agro-industrial waste material and used to compare with those the conventional support-porous material. The abundant of biomass wastes generated in palm oil industries is one potential source to convert the wastes into sustainable material as replacement of support material for catalysts. Objective of this study was to compare the kinetic rate of reaction the combustion of methane on Palladium (Pd) based catalyst with Al2O3 support and bio-char (Bc) support derived from shell kernel. The 2wt% Pd was prepared using incipient wetness impregnation method and the HCC performance was accomplished using tubular quartz reactor with gas mixture ratio of 3% methane and 97% air. Material characterization was determined using TGA, SEM, and BET surface area. The methane porous-HCC conversion was carried out by online gas analyzer connected to the reactor that performed porous-HCC. BET surface area for prepared 2 wt% Pd/Bc is smaller than prepared 2wt% Pd/ Al2O3 due to its low porosity between particles. The order of catalyst activity based on kinetic rate on reaction of catalysts in low temperature is prepared 2wt% Pd/Bc > calcined 2wt% Pd/ Al2O3 > prepared 2wt% Pd/ Al2O3 > calcined 2wt% Pd/Bc. Hence the usage of agro-industrial bio-mass waste material can enhance the sustainability principle.

Keywords: catalytic-combustion, environmental, support-bio-char material, sustainable and renewable material

Procedia PDF Downloads 369

Authors: Rohit T. P., Sahil Athrij, Sasi Gopalan

Abstract:

Currently, the content entertainment industry is dominated by mobile devices. As the trends slowly shift towards Augmented/Virtual Reality applications the computational demands on these devices are increasing exponentially and we are already reaching the limits of hardware optimizations. This paper proposes a software solution to this problem. By leveraging the capabilities of cloud computing we can offload the work from mobile devices to dedicated rendering servers that are way more powerful. But this introduces the problem of latency. This paper introduces a protocol that can achieve high-performance low latency Augmented/Virtual Reality experience. There are two parts to the protocol, 1) In-flight compression The main cause of latency in the system is the time required to transmit the camera frame from client to server. The round trip time is directly proportional to the amount of data transmitted. This can therefore be reduced by compressing the frames before sending. Using some standard compression algorithms like JPEG can result in minor size reduction only. Since the images to be compressed are consecutive camera frames there won't be a lot of changes between two consecutive images. So inter-frame compression is preferred. Inter-frame compression can be implemented efficiently using WebGL but the implementation of WebGL limits the precision of floating point numbers to 16bit in most devices. This can introduce noise to the image due to rounding errors, which will add up eventually. This can be solved using an improved interframe compression algorithm. The algorithm detects changes between frames and reuses unchanged pixels from the previous frame. This eliminates the need for floating point subtraction thereby cutting down on noise. The change detection is also improved drastically by taking the weighted average difference of pixels instead of the absolute difference. The kernel weights for this comparison can be fine-tuned to match the type of image to be compressed. 2) Dynamic Load distribution Conventional cloud computing architectures work by offloading as much work as possible to the servers, but this approach can cause a hit on bandwidth and server costs. The most optimal solution is obtained when the device utilizes 100% of its resources and the rest is done by the server. The protocol balances the load between the server and the client by doing a fraction of the computing on the device depending on the power of the device and network conditions. The protocol will be responsible for dynamically partitioning the tasks. Special flags will be used to communicate the workload fraction between the client and the server and will be updated in a constant interval of time ( or frames ). The whole of the protocol is designed so that it can be client agnostic. Flags are available to the client for resetting the frame, indicating latency, switching mode, etc. The server can react to client-side changes on the fly and adapt accordingly by switching to different pipelines. The server is designed to effectively spread the load and thereby scale horizontally. This is achieved by isolating client connections into different processes.

Keywords: 2D kernelling, augmented reality, cloud computing, dynamic load distribution, immersive experience, mobile computing, motion tracking, protocols, real-time systems, web-based augmented reality application

Procedia PDF Downloads 48

Authors: V. Saccà, A. Sarica, F. Novellino, S. Barone, T. Tallarico, E. Filippelli, A. Granata, P. Valentino, A. Quattrone

Abstract:

The work aim was to evaluate how well Random Forest (RF) and Support Vector Machine (SVM) algorithms could support the early diagnosis of Multiple Sclerosis (MS) from resting-state functional connectivity data. In particular, we wanted to explore the ability in distinguishing between controls and patients of mean signals extracted from ICA components corresponding to 15 well-known networks. Eighteen patients with early-MS (mean-age 37.42±8.11, 9 females) were recruited according to McDonald and Polman, and matched for demographic variables with 19 healthy controls (mean-age 37.55±14.76, 10 females). MRI was acquired by a 3T scanner with 8-channel head coil: (a)whole-brain T1-weighted; (b)conventional T2-weighted; (c)resting-state functional MRI (rsFMRI), 200 volumes. Estimated total lesion load (ml) and number of lesions were calculated using LST-toolbox from the corrected T1 and FLAIR. All rsFMRIs were pre-processed using tools from the FMRIB's Software Library as follows: (1) discarding of the first 5 volumes to remove T1 equilibrium effects, (2) skull-stripping of images, (3) motion and slice-time correction, (4) denoising with high-pass temporal filter (128s), (5) spatial smoothing with a Gaussian kernel of FWHM 8mm. No statistical significant differences (t-test, p < 0.05) were found between the two groups in the mean Euclidian distance and the mean Euler angle. WM and CSF signal together with 6 motion parameters were regressed out from the time series. We applied an independent component analysis (ICA) with the GIFT-toolbox using the Infomax approach with number of components=21. Fifteen mean components were visually identified by two experts. The resulting z-score maps were thresholded and binarized to extract the mean signal of the 15 networks for each subject. Statistical and machine learning analysis were then conducted on this dataset composed of 37 rows (subjects) and 15 features (mean signal in the network) with R language. The dataset was randomly splitted into training (75%) and test sets and two different classifiers were trained: RF and RBF-SVM. We used the intrinsic feature selection of RF, based on the Gini index, and recursive feature elimination (rfe) for the SVM, to obtain a rank of the most predictive variables. Thus, we built two new classifiers only on the most important features and we evaluated the accuracies (with and without feature selection) on test-set. The classifiers, trained on all the features, showed very poor accuracies on training (RF:58.62%, SVM:65.52%) and test sets (RF:62.5%, SVM:50%). Interestingly, when feature selection by RF and rfe-SVM were performed, the most important variable was the sensori-motor network I in both cases. Indeed, with only this network, RF and SVM classifiers reached an accuracy of 87.5% on test-set. More interestingly, the only misclassified patient resulted to have the lowest value of lesion volume. We showed that, with two different classification algorithms and feature selection approaches, the best discriminant network between controls and early MS, was the sensori-motor I. Similar importance values were obtained for the sensori-motor II, cerebellum and working memory networks. These findings, in according to the early manifestation of motor/sensorial deficits in MS, could represent an encouraging step toward the translation to the clinical diagnosis and prognosis.

Keywords: feature selection, machine learning, multiple sclerosis, random forest, support vector machine

Procedia PDF Downloads 215

Authors: Linna Li

Abstract:

The capability to move around places is doubtless very important for individuals to maintain good health and social functions. People’s activities in space and time have long been a research topic in behavioral and socio-economic studies, particularly focusing on the highly dynamic urban environment. By analyzing groups of people who share similar activity patterns, many socio-economic and socio-demographic problems and their relationships with individual behavior preferences can be revealed. Los Angeles, known for its large population, ethnic diversity, cultural mixing, and entertainment industry, faces great transportation challenges such as traffic congestion, parking difficulties, and long commuting. Understanding people’s travel behavior and movement patterns in this metropolis sheds light on potential solutions to complex problems regarding urban mobility. This project visualizes people’s trajectories in Greater Los Angeles (L.A.) Area over a period of two months using Twitter data. A Python script was used to collect georeferenced tweets within the Greater L.A. Area including Ventura, San Bernardino, Riverside, Los Angeles, and Orange counties. Information associated with tweets includes text, time, location, and user ID. Information associated with users includes name, the number of followers, etc. Both aggregated and individual activity patterns are demonstrated using various geovisualization techniques. Locations of individual Twitter users were aggregated to create a surface of activity hot spots at different time instants using kernel density estimation, which shows the dynamic flow of people’s movement throughout the metropolis in a twenty-four-hour cycle. In the 3D geovisualization interface, the z-axis indicates time that covers 24 hours, and the x-y plane shows the geographic space of the city. Any two points on the z axis can be selected for displaying activity density surface within a particular time period. In addition, daily trajectories of Twitter users were created using space-time paths that show the continuous movement of individuals throughout the day. When a personal trajectory is overlaid on top of ancillary layers including land use and road networks in 3D visualization, the vivid representation of a realistic view of the urban environment boosts situational awareness of the map reader. A comparison of the same individual’s paths on different days shows some regular patterns on weekdays for some Twitter users, but for some other users, their daily trajectories are more irregular and sporadic. This research makes contributions in two major areas: geovisualization of spatial footprints to understand travel behavior using the big data approach and dynamic representation of activity space in the Greater Los Angeles Area. Unlike traditional travel surveys, social media (e.g., Twitter) provides an inexpensive way of data collection on spatio-temporal footprints. The visualization techniques used in this project are also valuable for analyzing other spatio-temporal data in the exploratory stage, thus leading to informed decisions about generating and testing hypotheses for further investigation. The next step of this research is to separate users into different groups based on gender/ethnic origin and compare their daily trajectory patterns.

Keywords: geovisualization, human mobility pattern, Los Angeles, social media

Procedia PDF Downloads 89

Authors: Heather E. Ewart, Keith Jensen, Rebecca N. Cliffe

Abstract:

Fragmentation and changes in the structural composition of tropical forests – as a result of intensifying anthropogenic disturbance – are increasing pressures on local biodiversity. Species with low dispersal abilities have some of the highest extinction risks in response to environmental change, as even small-scale environmental variation can substantially impact their space use and energetic balance. Understanding the implications of forest disturbance is therefore essential, ultimately allowing for more effective and targeted conservation initiatives. Here, the impact of different levels of forest disturbance on the space use, energetics, movement and behavior of 18 brown-throated sloths (Bradypus variegatus) were assessed in the South Caribbean of Costa Rica. A multi-scale framework was used to measure forest disturbance, including large-scale (landscape-level classifications) and fine-scale (within and surrounding individual home ranges) forest composition. Three landscape-level classifications were identified: primary forests (undisturbed), secondary forests (some disturbance, regenerating) and urban forests (high levels of disturbance and fragmentation). Finer-scale forest composition was determined using measurements of habitat structure and quality within and surrounding individual home ranges for each sloth (home range estimates were calculated using autocorrelated kernel density estimation [AKDE]). Measurements of forest quality included tree connectivity, density, diameter and height, species richness, and percentage of canopy cover. To determine space use, energetics, movement and behavior, six sloths in urban forests, seven sloths in secondary forests and five sloths in primary forests were tracked using a combination of Very High Frequency (VHF) radio transmitters and Global Positioning System (GPS) technology over an average period of 120 days. All sloths were also fitted with micro data-loggers (containing tri-axial accelerometers and pressure loggers) for an average of 30 days to allow for behavior-specific movement analyses (data analysis ongoing for data-loggers and primary forest sloths). Data-loggers included determination of activity budgets, circadian rhythms of activity and energy expenditure (using the vector of the dynamic body acceleration [VeDBA] as a proxy). Analyses to date indicate that home range size significantly increased with the level of forest disturbance. Female sloths inhabiting secondary forests averaged 0.67-hectare home ranges, while female sloths inhabiting urban forests averaged 1.93-hectare home ranges (estimates are represented by median values to account for the individual variation in home range size in sloths). Likewise, home range estimates for male sloths were 2.35 hectares in secondary forests and 4.83 in urban forests. Sloths in urban forests also used nearly double (median = 22.5) the number of trees as sloths in the secondary forest (median = 12). These preliminary data indicate that forest disturbance likely heightens the energetic requirements of sloths, a species already critically limited by low dispersal ability and rates of energy acquisition. Energetic and behavioral analyses from the data-loggers will be considered in the context of fine-scale forest composition measurements (i.e., habitat quality and structure) and are expected to reflect the observed home range and movement constraints. The implications of these results are far-reaching, presenting an opportunity to define a critical index of habitat connectivity for low dispersal species such as sloths.

Keywords: biodiversity conservation, forest disturbance, movement ecology, sloths

Procedia PDF Downloads 70