Search results for: LiDAR datasets

Authors: Rodolfo Lorbieski, Silvia Modesto Nassar

Abstract:

Many supervised machine learning tasks require decision making across numerous different classes. Multi-class classification has several applications, such as face recognition, text recognition and medical diagnostics. The objective of this article is to analyze an adapted method of Stacking in multi-class problems, which combines ensembles within the ensemble itself. For this purpose, a training similar to Stacking was used, but with three levels, where the final decision-maker (level 2) performs its training by combining outputs from the tree-based pair of meta-classifiers (level 1) from Bayesian families. These are in turn trained by pairs of base classifiers (level 0) of the same family. This strategy seeks to promote diversity among the ensembles forming the meta-classifier level 2. Three performance measures were used: (1) accuracy, (2) area under the ROC curve, and (3) time for three factors: (a) datasets, (b) experiments and (c) levels. To compare the factors, ANOVA three-way test was executed for each performance measure, considering 5 datasets by 25 experiments by 3 levels. A triple interaction between factors was observed only in time. The accuracy and area under the ROC curve presented similar results, showing a double interaction between level and experiment, as well as for the dataset factor. It was concluded that level 2 had an average performance above the other levels and that the proposed method is especially efficient for multi-class problems when compared to binary problems.

Keywords: stacking, multi-layers, ensemble, multi-class

Procedia PDF Downloads 269

Authors: Taysir Soliman

Abstract:

One of the current popular clustering techniques is Spectral Clustering (SC) because of its advantages over conventional approaches such as hierarchical clustering, k-means, etc. and other techniques as well. However, one of the disadvantages of SC is the time consuming process because it requires computing the eigenvectors. In the past to overcome this disadvantage, a number of attempts have been proposed such as the Power Iteration Clustering (PIC) technique, which is one of versions from SC; some of PIC advantages are: 1) its scalability and efficiency, 2) finding one pseudo-eigenvectors instead of computing eigenvectors, and 3) linear combination of the eigenvectors in linear time. However, its worst disadvantage is an inter-class collision problem because it used only one pseudo-eigenvectors which is not enough. Previous researchers developed Deflation-based Power Iteration Clustering (DPIC) to overcome problems of PIC technique on inter-class collision with the same efficiency of PIC. In this paper, we developed Parallel DPIC (PDPIC) to improve the time and memory complexity which is run on apache spark framework using sparse matrix. To test the performance of PDPIC, we compared it to SC, ESCG, ESCALG algorithms on four small graph benchmark datasets and nine large graph benchmark datasets, where PDPIC proved higher accuracy and better time consuming than other compared algorithms.

Keywords: spectral clustering, power iteration clustering, deflation-based power iteration clustering, Apache spark, large graph

Procedia PDF Downloads 189

Authors: R. Nandhini, Gaurab Mudbhari

Abstract:

Ranging from the field of health care to self-driving cars, machine learning and deep learning algorithms have revolutionized the field with the proper utilization of images and visual-oriented data. Segmentation, regression, classification, clustering, dimensionality reduction, etc., are some of the Machine Learning tasks that helped Machine Learning and Deep Learning models to become state-of-the-art models for the field where images are key datasets. Among these tasks, classification and clustering are essential but difficult because of the intricate and high-dimensional characteristics of image data. This finding examines and assesses advanced techniques in supervised classification and unsupervised clustering for image datasets, emphasizing the relative efficiency of Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), Deep Embedded Clustering (DEC), and self-supervised learning approaches. Due to the distinctive structural attributes present in images, conventional methods often fail to effectively capture spatial patterns, resulting in the development of models that utilize more advanced architectures and attention mechanisms. In image classification, we investigated both CNNs and ViTs. One of the most promising models, which is very much known for its ability to detect spatial hierarchies, is CNN, and it serves as a core model in our study. On the other hand, ViT is another model that also serves as a core model, reflecting a modern classification method that uses a self-attention mechanism which makes them more robust as this self-attention mechanism allows them to lean global dependencies in images without relying on convolutional layers. This paper evaluates the performance of these two architectures based on accuracy, precision, recall, and F1-score across different image datasets, analyzing their appropriateness for various categories of images. In the domain of clustering, we assess DEC, Variational Autoencoders (VAEs), and conventional clustering techniques like k-means, which are used on embeddings derived from CNN models. DEC, a prominent model in the field of clustering, has gained the attention of many ML engineers because of its ability to combine feature learning and clustering into a single framework and its main goal is to improve clustering quality through better feature representation. VAEs, on the other hand, are pretty well known for using latent embeddings for grouping similar images without requiring for prior label by utilizing the probabilistic clustering method.

Keywords: machine learning, deep learning, image classification, image clustering

Procedia PDF Downloads 10

Authors: N. A. Parvin, E. B. Ferranti, L. A. Chapman, C. A. Pfrang

Abstract:

This study intends to investigate the effects of the Urban Heat Island on public health in Birmingham. Birmingham is located at the center of the West Midlands and its weather is Highly variable due to geographical factors. Residential developments, road networks and infrastructure often replace open spaces and vegetation. This transformation causes the temperature of urban areas to increase and creates an "island" of higher temperatures in the urban landscape. Extreme heat in the urban area is influencing public health in the UK as well as in the world. Birmingham is a densely built-up area with skyscrapers and congested buildings in the city center, which is a barrier to air circulation. We will investigate the city regarding heat and cold-related human mortality and other impacts. We are using primary and secondary datasets to examine the effect of population shift and land-use change on the UHI in Birmingham. We will also use freely available weather data from the Birmingham Urban Observatory and will incorporate satellite data to determine urban spatial expansion and its effect on the UHI. We have produced a temperature map based on summer datasets of 2020, which has covered 25 weather stations in Birmingham to show the differences between diurnal and nocturnal summer and annual temperature trends. Some impacts of the UHI may be beneficial, such as the lengthening of the plant growing season, but most of them are highly negative. We are looking for various effects of urban heat which is impacting human health and investigating mitigation options.

Keywords: urban heat, public health, climate change

Procedia PDF Downloads 96

Authors: Daniel R. dos Santos, Mateus S. Maldonado, Estevão J. R. Batista

Abstract:

The military commanders increasingly dependent on spatial awareness, as knowing where enemy are, understanding how war battle scenarios change over time, and visualizing these trends in ways that offer insights for decision-making. Thanks to advancements in geospatial technologies and artificial intelligence algorithms, the commanders are now able to modernize military operations on a universal scale. Thus, geovisualisation has become an essential asset in the defense sector. It has become indispensable for better decisionmaking in dynamic/temporal scenarios, operation planning and management for the war field, situational awareness, effective planning, monitoring, and others. For example, a 3D visualization of war field data contributes to intelligence analysis, evaluation of postmission outcomes, and creation of predictive models to enhance decision-making and strategic planning capabilities. However, old-school visualization methods are slow, expensive, and unscalable. Despite modern technologies in generating 3D point clouds, such as LIDAR and stereo sensors, monocular depth values based on deep learning can offer a faster and more detailed view of the environment, transforming single images into visual information for valuable insights. We propose a dedicated monocular depth reconstruction approach via deep learning techniques for 3D geovisualisation of satellite images. It introduces scalability in terrain reconstruction and data visualization. First, a dataset with more than 7,000 satellite images and associated digital elevation model (DEM) is created. It is based on high resolution optical and radar imageries collected from Planet and Copernicus, on which we fuse highresolution topographic data obtained using technologies such as LiDAR and the associated geographic coordinates. Second, we developed an imagery-DEM fusion strategy that combine feature maps from two encoder-decoder networks. One network is trained with radar and optical bands, while the other is trained with DEM features to compute dense 3D depth. Finally, we constructed a benchmark with sparse depth annotations to facilitate future research. To demonstrate the proposed method's versatility, we evaluated its performance on no annotated satellite images and implemented an enclosed environment useful for Geovisualisation applications. The algorithms were developed in Python 3.0, employing open-source computing libraries, i.e., Open3D, TensorFlow, and Pythorch3D. The proposed method provides fast and accurate decision-making with GIS for localization of troops, position of the enemy, terrain and climate conditions. This analysis enhances situational consciousness, enabling commanders to fine-tune the strategies and distribute the resources proficiently.

Keywords: depth, deep learning, geovisualisation, satellite images

Procedia PDF Downloads 10

Authors: Ki-Yeo Kim

Abstract:

Trends in genetics are transforming in order to identify differential coexpressions of correlated gene expression rather than the significant individual gene. Moreover, it is known that a combined biomarker pattern improves the discrimination of a specific cancer. The identification of the combined biomarker is also necessary for the early detection of invasive oral squamous cell carcinoma (OSCC). To identify the combined biomarker that could improve the discrimination of OSCC, we explored an appropriate number of genes in a combined gene set in order to attain the highest level of accuracy. After detecting a significant gene set, including the pre-defined number of genes, a combined expression was identified using the weights of genes in a gene set. We used the Principal Component Analysis (PCA) for the weight calculation. In this process, we used three public microarray datasets. One dataset was used for identifying the combined biomarker, and the other two datasets were used for validation. The discrimination accuracy was measured by the out-of-bag (OOB) error. There was no relation between the significance and the discrimination accuracy in each individual gene. The identified gene set included both significant and insignificant genes. One of the most significant gene sets in the classification of normal and OSCC included MMP1, SOCS3 and ACOX1. Furthermore, in the case of oral dysplasia and OSCC discrimination, two combined biomarkers were identified. The combined genomic expression achieved better performance in the discrimination of different conditions than in a single significant gene. Therefore, it could be expected that accurate diagnosis for cancer could be possible with a combined biomarker.

Keywords: oral squamous cell carcinoma, combined biomarker, microarray dataset, correlated genes

Procedia PDF Downloads 423

Authors: Sanjay Rathee, Arti Kashyap

Abstract:

Extraction of useful information from large datasets is one of the most important research problems. Association rule mining is one of the best methods for this purpose. Finding possible associations between items in large transaction based datasets (finding frequent patterns) is most important part of the association rule mining. There exist many algorithms to find frequent patterns but Apriori algorithm always remains a preferred choice due to its ease of implementation and natural tendency to be parallelized. Many single-machine based Apriori variants exist but massive amount of data available these days is above capacity of a single machine. Therefore, to meet the demands of this ever-growing huge data, there is a need of multiple machines based Apriori algorithm. For these types of distributed applications, MapReduce is a popular fault-tolerant framework. Hadoop is one of the best open-source software frameworks with MapReduce approach for distributed storage and distributed processing of huge datasets using clusters built from commodity hardware. However, heavy disk I/O operation at each iteration of a highly iterative algorithm like Apriori makes Hadoop inefficient. A number of MapReduce-based platforms are being developed for parallel computing in recent years. Among them, two platforms, namely, Spark and Flink have attracted a lot of attention because of their inbuilt support to distributed computations. Earlier we proposed a reduced- Apriori algorithm on Spark platform which outperforms parallel Apriori, one because of use of Spark and secondly because of the improvement we proposed in standard Apriori. Therefore, this work is a natural sequel of our work and targets on implementing, testing and benchmarking Apriori and Reduced-Apriori and our new algorithm ReducedAll-Apriori on Apache Flink and compares it with Spark implementation. Flink, a streaming dataflow engine, overcomes disk I/O bottlenecks in MapReduce, providing an ideal platform for distributed Apriori. Flink's pipelining based structure allows starting a next iteration as soon as partial results of earlier iteration are available. Therefore, there is no need to wait for all reducers result to start a next iteration. We conduct in-depth experiments to gain insight into the effectiveness, efficiency and scalability of the Apriori and RA-Apriori algorithm on Flink.

Keywords: apriori, apache flink, Mapreduce, spark, Hadoop, R-Apriori, frequent itemset mining

Procedia PDF Downloads 294

Authors: Ahmed Elaksher, Islam Omar

Abstract:

For many years, Mars surface has been a mystery for scientists. Lately with the help of geospatial data and photogrammetric procedures researchers were able to capture some insights about this planet. Two of the most imperative data sources to explore Mars are the The High Resolution Imaging Science Experiment (HiRISE) and the Mars Orbiter Laser Altimeter (MOLA). HiRISE is one of six science instruments carried by the Mars Reconnaissance Orbiter, launched August 12, 2005, and managed by NASA. The MOLA sensor is a laser altimeter carried by the Mars Global Surveyor (MGS) and launched on November 7, 1996. In this project, we used MOLA-based DEMs to orthorectify HiRISE optical images for generating a more accurate and trustful surface of Mars. The MOLA data was interpolated using the kriging interpolation technique. Corresponding tie points were digitized from both datasets. These points were employed in co-registering both datasets using GIS analysis tools. In this project, we employed three different 3D to 2D transformation models. These are the parallel projection (3D affine) transformation model; the extended parallel projection transformation model; the Direct Linear Transformation (DLT) model. A set of tie-points was digitized from both datasets. These points were split into two sets: Ground Control Points (GCPs), used to evaluate the transformation parameters using least squares adjustment techniques, and check points (ChkPs) to evaluate the computed transformation parameters. Results were evaluated using the RMSEs between the precise horizontal coordinates of the digitized check points and those estimated through the transformation models using the computed transformation parameters. For each set of GCPs, three different configurations of GCPs and check points were tested, and average RMSEs are reported. It was found that for the 2D transformation models, average RMSEs were in the range of five meters. Increasing the number of GCPs from six to ten points improve the accuracy of the results with about two and half meters. Further increasing the number of GCPs didn’t improve the results significantly. Using the 3D to 2D transformation parameters provided three to two meters accuracy. Best results were reported using the DLT transformation model. However, increasing the number of GCPS didn’t have substantial effect. The results support the use of the DLT model as it provides the required accuracy for ASPRS large scale mapping standards. However, well distributed sets of GCPs is a key to provide such accuracy. The model is simple to apply and doesn’t need substantial computations.

Keywords: mars, photogrammetry, MOLA, HiRISE

Procedia PDF Downloads 57

Authors: Nikolaos Vassilas, Theocharis Tsenoglou, Djamchid Ghazanfarpour

Abstract:

In this work we explore the capability of the mean shift algorithm as a powerful preprocessing tool for improving the quality of spatial data, acquired from airborne scanners, from densely built urban areas. On one hand, high resolution image data corrupted by noise caused by lossy compression techniques are appropriately smoothed while at the same time preserving the optical edges and, on the other, low resolution LiDAR data in the form of normalized Digital Surface Map (nDSM) is upsampled through the joint mean shift algorithm. Experiments on both the edge-preserving smoothing and upsampling capabilities using synthetic RGB-z data show that the mean shift algorithm is superior to bilateral filtering as well as to other classical smoothing and upsampling algorithms. Application of the proposed methodology for 3D reconstruction of buildings of a pilot region of Athens, Greece results in a significant visual improvement of the 3D building block model.

Keywords: 3D buildings reconstruction, data fusion, data upsampling, mean shift

Procedia PDF Downloads 315

Authors: Ali Akyar, Osman Serdar Gedik

Abstract:

Depth estimation is a challenging computer vision task that involves estimating the distance between objects in a scene and the camera. It predicts how far each pixel in the 2D image is from the capturing point. There are some important Monocular Depth Estimation (MDE) studies that are based on Vision Transformers (ViT). We benchmark three major studies. The first work aims to build a simple and powerful foundation model that deals with any images under any condition. The second work proposes a method by mixing multiple datasets during training and a robust training objective. The third work combines generalization performance and state-of-the-art results on specific datasets. Although there are studies with thermal images too, we wanted to benchmark these three non-thermal, state-of-the-art studies with a hybrid image dataset which is taken by Multi-Spectral Dynamic Imaging (MSX) technology. MSX technology produces detailed thermal images by bringing together the thermal and visual spectrums. Using this technology, our dataset images are not blur and poorly detailed as the normal thermal images. On the other hand, they are not taken at the perfect light conditions as RGB images. We compared three methods under test with our thermal dataset which was not done before. Additionally, we propose an image enhancement deep learning model for thermal data. This model helps extract the features required for monocular depth estimation. The experimental results demonstrate that, after using our proposed model, the performance of these three methods under test increased significantly for thermal image depth prediction.

Keywords: monocular depth estimation, thermal dataset, benchmarking, vision transformers

Procedia PDF Downloads 32

Authors: Ali Almadan, Anoop Krishnan, Ajita Rattani

Abstract:

Facial recognition is a widely adopted technology in surveillance, border control, healthcare, banking services, and lately, in mobile user authentication with Apple introducing “Face ID” moniker with iPhone X. A lot of research has been conducted in the area of face recognition on datasets captured by surveillance cameras, DSLR, and mobile devices. Recently, face recognition technology has also been deployed on body-worn cameras to keep officers safe, enabling situational awareness and providing evidence for trial. However, limited academic research has been conducted on this topic so far, without the availability of any publicly available datasets with a sufficient sample size. This paper aims to advance research in the area of face recognition using body-worn cameras. To this aim, the contribution of this work is two-fold: (1) collection of a dataset consisting of a total of 136,939 facial images of 102 subjects captured using body-worn cameras in in-door and daylight conditions and (2) evaluation of various deep-learning architectures for face identification on the collected dataset. Experimental results suggest a maximum True Positive Rate(TPR) of 99.86% at False Positive Rate(FPR) of 0.000 obtained by SphereFace based deep learning architecture in daylight condition. The collected dataset and the baseline algorithms will promote further research and development. A downloadable link of the dataset and the algorithms is available by contacting the authors.

Keywords: face recognition, body-worn cameras, deep learning, person identification

Procedia PDF Downloads 163

Authors: Busayo Fashoto, Abdulhakeem Shaibu, Justice Agbadu, Samuel Aiyeoribe

Abstract:

Data has gone from just rows and columns to being an infrastructure itself. The traditional medium of data infrastructure has been managed by individuals in different industries and saved on personal work tools; one of such is the laptop. This hinders data sharing and Sustainable Development Goal (SDG) 9 for infrastructure sustainability across all countries and regions. However, there has been a constant demand for data across different agencies and ministries by investors and decision-makers. The rapid development and adoption of open-source technologies that promote the collection and processing of data in new ways and in ever-increasing volumes are creating new data infrastructure in sectors such as lands and health, among others. This paper examines the process of developing data infrastructure and, by extension, a data portal to provide baseline data for sustainable development and decision making in Nigeria. This paper employs the FAIR principle (Findable, Accessible, Interoperable, and Reusable) of data management using open-source technology tools to develop data portals for public use. eHealth Africa, an organization that uses technology to drive public health interventions in Nigeria, developed a data portal which is a typical data infrastructure that serves as a repository for various datasets on administrative boundaries, points of interest, settlements, social infrastructure, amenities, and others. This portal makes it possible for users to have access to datasets of interest at any point in time at no cost. A skeletal infrastructure of this data portal encompasses the use of open-source technology such as Postgres database, GeoServer, GeoNetwork, and CKan. These tools made the infrastructure sustainable, thus promoting the achievement of SDG 9 (Industries, Innovation, and Infrastructure). As of 6th August 2021, a wider cross-section of 8192 users had been created, 2262 datasets had been downloaded, and 817 maps had been created from the platform. This paper shows the use of rapid development and adoption of technologies that facilitates data collection, processing, and publishing in new ways and in ever-increasing volumes. In addition, the paper is explicit on new data infrastructure in sectors such as health, social amenities, and agriculture. Furthermore, this paper reveals the importance of cross-sectional data infrastructures for planning and decision making, which in turn can form a central data repository for sustainable development across developing countries.

Keywords: data portal, data infrastructure, open source, sustainability

Procedia PDF Downloads 97

Authors: James Jr. Mashiyane, Risuna Nkolele, Stephanie J. Müller, Gciniwe S. Dlamini, Rebone L. Meraba, Darlington S. Mapiye

Abstract:

When performing language tasks in natural language processing (NLP), the dimensionality of word embeddings is chosen either ad-hoc or is calculated by optimizing the Pairwise Inner Product (PIP) loss. The PIP loss is a metric that measures the dissimilarity between word embeddings, and it is obtained through matrix perturbation theory by utilizing the unitary invariance of word embeddings. Unlike in natural language, in genomics, especially in genome sequence processing, unlike in natural language processing, there is no notion of a “word,” but rather, there are sequence substrings of length k called k-mers. K-mers sizes matter, and they vary depending on the goal of the task at hand. The dimensionality of word embeddings in NLP has been studied using the matrix perturbation theory and the PIP loss. In this paper, the sufficiency and reliability of applying word-embedding algorithms to various genomic sequence datasets are investigated to understand the relationship between the k-mer size and their embedding dimension. This is completed by studying the scaling capability of three embedding algorithms, namely Latent Semantic analysis (LSA), Word2Vec, and Global Vectors (GloVe), with respect to the k-mer size. Utilising the PIP loss as a metric to train embeddings on different datasets, we also show that Word2Vec outperforms LSA and GloVe in accurate computing embeddings as both the k-mer size and vocabulary increase. Finally, the shortcomings of natural language processing embedding algorithms in performing genomic tasks are discussed.

Keywords: word embeddings, k-mer embedding, dimensionality reduction

Procedia PDF Downloads 137

Authors: Sara ElElimy, Samir Moustafa

Abstract:

Mobile network operators start to face many challenges in the digital era, especially with high demands from customers. Since mobile network operators are considered a source of big data, traditional techniques are not effective with new era of big data, Internet of things (IoT) and 5G; as a result, handling effectively different big datasets becomes a vital task for operators with the continuous growth of data and moving from long term evolution (LTE) to 5G. So, there is an urgent need for effective Big data analytics to predict future demands, traffic, and network performance to full fill the requirements of the fifth generation of mobile network technology. In this paper, we introduce data science techniques using machine learning and deep learning algorithms: the autoregressive integrated moving average (ARIMA), Bayesian-based curve fitting, and recurrent neural network (RNN) are employed for a data-driven application to mobile network operators. The main framework included in models are identification parameters of each model, estimation, prediction, and final data-driven application of this prediction from business and network performance applications. These models are applied to Telecom Italia Big Data challenge call detail records (CDRs) datasets. The performance of these models is found out using a specific well-known evaluation criteria shows that ARIMA (machine learning-based model) is more accurate as a predictive model in such a dataset than the RNN (deep learning model).

Keywords: big data analytics, machine learning, CDRs, 5G

Procedia PDF Downloads 139

Authors: Chen Guo, Heng Tang, Ben Niu

Abstract:

Clustering splits objects into different groups based on similarity, making the objects have higher similarity in the same group and lower similarity in different groups. Thus, clustering can be treated as an optimization problem to maximize the intra-cluster similarity or inter-cluster dissimilarity. In real-world applications, the datasets often have some complex characteristics: sparse, overlap, high dimensionality, etc. When facing these datasets, simultaneously optimizing two or more objectives can obtain better clustering results than optimizing one objective. However, except for the objectives weighting methods, traditional clustering approaches have difficulty in solving multi-objective data clustering problems. Due to this, evolutionary multi-objective optimization algorithms are investigated by researchers to optimize multiple clustering objectives. In this paper, the Data Clustering algorithm based on Multi-objective Periodic Bacterial Foraging Optimization with two Learning Archives (DC-MPBFOLA) is proposed. Specifically, first, to reduce the high computing complexity of the original BFO, periodic BFO is employed as the basic algorithmic framework. Then transfer the periodic BFO into a multi-objective type. Second, two learning strategies are proposed based on the two learning archives to guide the bacterial swarm to move in a better direction. On the one hand, the global best is selected from the global learning archive according to the convergence index and diversity index. On the other hand, the personal best is selected from the personal learning archive according to the sum of weighted objectives. According to the aforementioned learning strategies, a chemotaxis operation is designed. Third, an elite learning strategy is designed to provide fresh power to the objects in two learning archives. When the objects in these two archives do not change for two consecutive times, randomly initializing one dimension of objects can prevent the proposed algorithm from falling into local optima. Fourth, to validate the performance of the proposed algorithm, DC-MPBFOLA is compared with four state-of-art evolutionary multi-objective optimization algorithms and one classical clustering algorithm on evaluation indexes of datasets. To further verify the effectiveness and feasibility of designed strategies in DC-MPBFOLA, variants of DC-MPBFOLA are also proposed. Experimental results demonstrate that DC-MPBFOLA outperforms its competitors regarding all evaluation indexes and clustering partitions. These results also indicate that the designed strategies positively influence the performance improvement of the original BFO.

Keywords: data clustering, multi-objective optimization, bacterial foraging optimization, learning archives

Procedia PDF Downloads 139

Authors: Sujit Kumar Pradhan, Anil Kumar, Vijay Kumar

Abstract:

The testing process of the software during the software development time is a crucial step as it makes the software more efficient and dependable. To estimate software’s reliability through the mean value function, many software reliability growth models (SRGMs) were developed under the assumption that operating and testing environments are the same. Practically, it is not true because when the software works in a natural field environment, the reliability of the software differs. This article discussed an SRGM comprising change-point and imperfect debugging in a simulated testing environment. Later on, we extended it in a multi-release direction. Initially, the software was released to the market with few features. According to the market’s demand, the software company upgraded the current version by adding new features as time passed. Therefore, we have proposed a generalized multi-release SRGM where change-point and imperfect debugging concepts have been addressed in a simulated testing environment. The failure-increasing rate concept has been adopted to determine the change point for each software release. Based on nine goodness-of-fit criteria, the proposed model is validated on two real datasets. The results demonstrate that the proposed model fits the datasets better. We have also discussed the optimal release time of the software through a cost model by assuming that the testing and debugging costs are time-dependent.

Keywords: software reliability growth models, non-homogeneous Poisson process, multi-release software, mean value function, change-point, environmental factors

Procedia PDF Downloads 74

Authors: Zunaira Zaman, David Bohannon

Abstract:

This study explores the application of machine learning (ML) for detecting security vulnerabilities in source code. The research aims to assist organizations with large application portfolios and limited security testing capabilities in prioritizing security activities. ML-based approaches offer benefits such as increased confidence scores, false positives and negatives tuning, and automated feedback. The initial approach using natural language processing techniques to extract features achieved 86% accuracy during the training phase but suffered from overfitting and performed poorly on unseen datasets during testing. To address these issues, the study proposes using the abstract syntax tree (AST) for Java and C++ codebases to capture code semantics and structure and generate path-context representations for each function. The Code2Vec model architecture is used to learn distributed representations of source code snippets for training a machine-learning classifier for vulnerability prediction. The study evaluates the performance of the proposed methodology using two datasets and compares the results with existing approaches. The Devign dataset yielded 60% accuracy in predicting vulnerable code snippets and helped resist overfitting, while the Juliet Test Suite predicted specific vulnerabilities such as OS-Command Injection, Cryptographic, and Cross-Site Scripting vulnerabilities. The Code2Vec model achieved 75% accuracy and a 98% recall rate in predicting OS-Command Injection vulnerabilities. The study concludes that even partial AST representations of source code can be useful for vulnerability prediction. The approach has the potential for automated intelligent analysis of source code, including vulnerability prediction on unseen source code. State-of-the-art models using natural language processing techniques and CNN models with ensemble modelling techniques did not generalize well on unseen data and faced overfitting issues. However, predicting vulnerabilities in source code using machine learning poses challenges such as high dimensionality and complexity of source code, imbalanced datasets, and identifying specific types of vulnerabilities. Future work will address these challenges and expand the scope of the research.

Keywords: code embeddings, neural networks, natural language processing, OS command injection, software security, code properties

Procedia PDF Downloads 106

Authors: Mohammad Abu-Shaira, Weishi Shi

Abstract:

Adaptive learning, a commonly employed solution to drift, involves updating predictive models online during their operation to react to concept drifts, thereby serving as a critical component and natural extension for online learning systems that learn incrementally from each example. This paper introduces LSTM-SCCM “Long Short-Term Memory Stream Cruise Control Method”, a drift adaptation-as-a-service framework for online learning. LSTM-SCCM automates drift adaptation through prompt detection, drift magnitude quantification, dynamic hyperparameter tuning, performing shortterm optimization and model recalibration for immediate adjustments, and, when necessary, conducting long-term model recalibration to ensure deeper enhancements in model performance. LSTM-SCCM is incorporated into a suite of cutting-edge online regression models, assessing their performance across various types of concept drift using diverse datasets with varying characteristics. The findings demonstrate that LSTM-SCCM represents a notable advancement in both model performance and efficacy in handling concept drift occurrences. LSTM-SCCM stands out as the sole framework adept at effectively tackling concept drifts within regression scenarios. Its proactive approach to drift adaptation distinguishes it from conventional reactive methods, which typically rely on retraining after significant degradation to model performance caused by drifts. Additionally, LSTM-SCCM employs an in-memory approach combined with the Self-Adjusting Memory (SAM) architecture to enhance real-time processing and adaptability. The framework incorporates variable thresholding techniques and does not assume any particular data distribution, making it an ideal choice for managing high-dimensional datasets and efficiently handling large-scale data. Our experiments, which include abrupt, incremental, and gradual drifts across both low- and high-dimensional datasets with varying noise levels, and applied to four state-of-the-art online regression models, demonstrate that LSTM-SCCM is versatile and effective, rendering it a valuable solution for online regression models to address concept drift.

Keywords: automated drift detection and adaptation, concept drift, hyperparameters optimization, online and adaptive learning, regression

Procedia PDF Downloads 11

Authors: Farhan Aljuaidi

Abstract:

The rapid expansion of Alkarj city, Saudi Arabia, towards the outlet of Wadi AlAin is critical for the planners and decision makers. Nowadays, two major projects such as Salman bin Abdulaziz University compound and new industrial area are developed in this flood prone area where no channels are clear and identified. The main contribution of this study is to divert the flow away from these vital projects by reconstructing new channels. To do so, Lidar data were used to generate contour lines for the actual elevation of the highways and local roads. These data were analyzed and compared to the contour lines derived from the topographical maps 1:50.000. The magnitude of the expected flow was estimated using Snyder's Model based on the morphometric data acquired by DEM of the catchment area. The results indicate that maximum discharge peak reaches 2694,3 m3/sec, the mean is 303,7 m3/sec and the minimum is 74,3 m3/sec. The runoff was estimated at 252,2. 610 m3/s, the mean is 41,5. 610 m3/s and the minimum is 12,4. 610 m3/s.

Keywords: Desert flood, Saudi Arabia, Snyder's Model, flow estimation

Procedia PDF Downloads 309

Authors: Mohamed Dakki, Genevieve Robin, Marie Suet, Abdeljebbar Qninba, Mohamed A. El Agbani, Asmâa Ouassou, Rhimou El Hamoumi, Hichem Azafzaf, Sami Rebah, Claudia Feltrup-Azafzaf, Nafouel Hamouda, Wed a.L. Ibrahim, Hosni H. Asran, Amr A. Elhady, Haitham Ibrahim, Khaled Etayeb, Essam Bouras, Almokhtar Saied, Ashrof Glidan, Bakar M. Habib, Mohamed S. Sayoud, Nadjiba Bendjedda, Laura Dami, Clemence Deschamps, Elie Gaget, Jean-Yves Mondain-Monval, Pierre Defos Du Rau

Abstract:

In biodiversity monitoring, large datasets are becoming more and more widely available and are increasingly used globally to estimate species trends and con- servation status. These large-scale datasets challenge existing statistical analysis methods, many of which are not adapted to their size, incompleteness and heterogeneity. The development of scalable methods to impute missing data in incomplete large-scale monitoring datasets is crucial to balance sampling in time or space and thus better inform conservation policies. We developed a new method based on penalized Poisson models to impute and analyse incomplete monitoring data in a large-scale framework. The method al- lows parameterization of (a) space and time factors, (b) the main effects of predic- tor covariates, as well as (c) space–time interactions. It also benefits from robust statistical and computational capability in large-scale settings. The method was tested extensively on both simulated and real-life waterbird data, with the findings revealing that it outperforms six existing methods in terms of missing data imputation errors. Applying the method to 16 waterbird species, we estimated their long-term trends for the first time at the entire North African scale, a region where monitoring data suffer from many gaps in space and time series. This new approach opens promising perspectives to increase the accuracy of species-abundance trend estimations. We made it freely available in the r package ‘lori’ (https://CRAN.R-project.org/package=lori) and recommend its use for large- scale count data, particularly in citizen science monitoring programmes.

Keywords: biodiversity monitoring, high-dimensional statistics, incomplete count data, missing data imputation, waterbird trends in North-Africa

Procedia PDF Downloads 156

Authors: Qasim Munye, Juned Islam, Haseeb Qureshi, Syed Jung

Abstract:

Lung nodules are commonly identified in computed tomography (CT) scans by experienced radiologists at a relatively late stage. Early diagnosis can greatly increase survival. We propose using a pix2pix conditional generative adversarial network to generate realistic images simulating early-stage lung nodule growth. We have applied deep images prior to 2341 slices from 895 computed tomography (CT) scans from the Lung Image Database Consortium (LIDC) dataset to generate pseudo-healthy medical images. From these images, 819 were chosen to train a pix2pix network. We observed that for most of the images, the pix2pix network was able to generate images where the nodule increased in size and intensity across epochs. To evaluate the images, 400 generated images were chosen at random and shown to a medical student beside their corresponding original image. Of these 400 generated images, 384 were defined as satisfactory - meaning they resembled a nodule and were visually similar to the corresponding image. We believe that this generated dataset could be used as training data for neural networks to detect lung nodules at an early stage or to improve the accuracy of such networks. This is particularly significant as datasets containing the growth of early-stage nodules are scarce. This project shows that the combination of deep image prior and generative models could potentially open the door to creating larger datasets than currently possible and has the potential to increase the accuracy of medical classification tasks.

Keywords: medical technology, artificial intelligence, radiology, lung cancer

Procedia PDF Downloads 68

Authors: Y. T. van Popta

Abstract:

This paper focuses on the never thoroughly examined yet in native relevant late medieval maritime cultural landscape of the former Zuiderzee (A.D. 1170-1932) in the center part of the Netherlands. Especially the northeastern part of the region, nowadays known as the Noordoostpolder, testifies of the dynamic battle of the Dutch against the water. This highly dynamic maritime region developed from a lake district into a sea and eventually into a polder. By linking physical and cognitive datasets from the Noordoostpol-der region in a spatial environment, new information on a late medieval maritime culture is brought to light, giving the opportunity to: (i) create a modern day perception on the late medieval maritime cultural landscape of the region and (ii) to underline the value of interdisciplinary and spatial research in maritime archaeology in general. Since the large scale reclamations of the region (A.D. 1932-1968), many remains have been discovered of a drowned and eroded late medieval maritime culture, represented by lost islands, drowned settlements, cultivated lands, shipwrecks and socio-economic networks. Recent archaeological research has proved the existence of this late medieval maritime culture by the discovery of the remains of the drowned settlement Fenehuysen (Veenhuizen) and its surroundings. The fact that this settlement and its cultivated surroundings remained hidden for so long proves that a large part of the maritime cultural landscape is ‘invisible’ and can only be found by extensive interdisciplinary research.

Keywords: drowned settlements, late middle ages, lost islands, maritime cultural landscape, the Netherlands

Procedia PDF Downloads 213

Authors: Yash Jain

Abstract:

The US Center for Disease Control has recommended wearing masks to slow the spread of the virus. The research uses a video feed from a camera to conduct real-time classifications of whether or not a human is correctly wearing a mask, incorrectly wearing a mask, or not wearing a mask at all. Utilizing two distinct datasets from the open-source website Kaggle, a mask detection network had been trained. The first dataset that was used to train the model was titled 'Face Mask Detection' on Kaggle, where the dataset was retrieved from and the second dataset was titled 'Face Mask Dataset, which provided the data in a (YOLO Format)' so that the TinyYoloV3 model could be trained. Based on the data from Kaggle, two machine learning models were implemented and trained: a Tiny YoloV3 Real-time model and a two-stage neural network classifier. The two-stage neural network classifier had a first step of identifying distinct faces within the image, and the second step was a classifier to detect the state of the mask on the face and whether it was worn correctly, incorrectly, or no mask at all. The TinyYoloV3 was used for the live feed as well as for a comparison standpoint against the previous two-stage classifier and was trained using the darknet neural network framework. The two-stage classifier attained a mean average precision (MAP) of 80%, while the model trained using TinyYoloV3 real-time detection had a mean average precision (MAP) of 59%. Overall, both models were able to correctly classify stages/scenarios of no mask, mask, and incorrectly worn masks.

Keywords: datasets, classifier, mask-detection, real-time, TinyYoloV3, two-stage neural network classifier

Procedia PDF Downloads 162

Authors: Fadwa Alaskar, Fang-Chieh Chou, Carlos Flores, Xiao-Yun Lu, Alexandre M. Bayen

Abstract:

This article proposes a perception model for enhancing the accuracy and stability of car-following control of a longitudinally automated truck. We applied a fusion-based tracking algorithm on measurements of a single preceding vehicle needed for car-following control. This algorithm fuses two types of data, radar and LiDAR data, to obtain more accurate and robust longitudinal perception of the subject vehicle in various weather conditions. The filter’s resulting signals are fed to the gap control algorithm at every tracking loop composed by a high-level gap control and lower acceleration tracking system. Several highway tests have been performed with two trucks. The tests show accurate and fast tracking of the target, which impacts on the gap control loop positively. The experiments also show the fulfilment of control design requirements, such as fast speed variations tracking and robust time gap following.

Keywords: object tracking, perception, sensor fusion, adaptive cruise control, cooperative adaptive cruise control

Procedia PDF Downloads 229

Authors: M. C. Akay, A. Aybakan, H. Temeltas

Abstract:

Aerial and ground robots have various advantages of usage in different missions. Aerial robots can move quickly and get a different sight of view of the area, but those vehicles cannot carry heavy payloads. On the other hand, unmanned ground vehicles (UGVs) are slow moving vehicles, since those can carry heavier payloads than unmanned aerial vehicles (UAVs). In this context, we investigate the performances of various Similarity Metrics to provide a common map for Heterogeneous Robot Team (HRT) in complex environments. Within the usage of Lidar Odometry and Octree Mapping technique, the local 3D maps of the environment are gathered.  In order to obtain a common map for HRT, informative theoretic similarity metrics are exploited. All types of these similarity metrics gave adequate as allowable simulation time and accurate results that can be used in different types of applications. For the heterogeneous multi robot team, those methods can be used to match different types of maps.

Keywords: common maps, heterogeneous robot team, map matching, informative theoretic similarity metrics

Procedia PDF Downloads 168

Authors: Xiaocong Liu, Huazhen Wang, Ting He, Xiaozheng Li, Weihan Zhang, Jian Chen

Abstract:

The utilization of electronic medical record (EMR) data to establish the disease diagnosis model has become an important research content of biomedical informatics. Deep learning can automatically extract features from the massive data, which brings about breakthroughs in the study of EMR data. The challenge is that deep learning lacks semantic knowledge, which leads to impracticability in medical science. This research proposes a method of incorporating lexical-semantic knowledge from abundant entities into a convolutional neural network (CNN) framework for pediatric disease diagnosis. Firstly, medical terms are vectorized into Lexical Semantic Vectors (LSV), which are concatenated with the embedded word vectors of word2vec to enrich the feature representation. Secondly, the semantic distribution of medical terms serves as Semantic Decision Guide (SDG) for the optimization of deep learning models. The study evaluate the performance of LSV-SDG-CNN model on four kinds of Chinese EMR datasets. Additionally, CNN, LSV-CNN, and SDG-CNN are designed as baseline models for comparison. The experimental results show that LSV-SDG-CNN model outperforms baseline models on four kinds of Chinese EMR datasets. The best configuration of the model yielded an F1 score of 86.20%. The results clearly demonstrate that CNN has been effectively guided and optimized by lexical-semantic knowledge, and LSV-SDG-CNN model improves the disease classification accuracy with a clear margin.

Keywords: convolutional neural network, electronic medical record, feature representation, lexical semantics, semantic decision

Procedia PDF Downloads 126

Authors: Armin Nabaei, M. Omair Ahmad, M. N. S. Swamy

Abstract:

Many facial expression and emotion recognition methods in the traditional approaches of using LDA, PCA, and EBGM have been proposed. In recent years deep learning models have provided a unique platform addressing by automatically extracting the features for the detection of facial expression and emotions. However, deep networks require large training datasets to extract automatic features effectively. In this work, we propose an efficient emotion detection algorithm using face images when only small datasets are available for training. We design a deep network whose feature extraction capability is enhanced by utilizing several parallel modules between the input and output of the network, each focusing on the extraction of different types of coarse features with fined grained details to break the symmetry of produced information. In fact, we leverage long range dependencies, which is one of the main drawback of CNNs. We develop this work by introducing a Dynamic Soft-Margin SoftMax.The conventional SoftMax suffers from reaching to gold labels very soon, which take the model to over-fitting. Because it’s not able to determine adequately discriminant feature vectors for some variant class labels. We reduced the risk of over-fitting by using a dynamic shape of input tensor instead of static in SoftMax layer with specifying a desired Soft- Margin. In fact, it acts as a controller to how hard the model should work to push dissimilar embedding vectors apart. For the proposed Categorical Loss, by the objective of compacting the same class labels and separating different class labels in the normalized log domain.We select penalty for those predictions with high divergence from ground-truth labels.So, we shorten correct feature vectors and enlarge false prediction tensors, it means we assign more weights for those classes with conjunction to each other (namely, “hard labels to learn”). By doing this work, we constrain the model to generate more discriminate feature vectors for variant class labels. Finally, for the proposed optimizer, our focus is on solving weak convergence of Adam optimizer for a non-convex problem. Our noteworthy optimizer is working by an alternative updating gradient procedure with an exponential weighted moving average function for faster convergence and exploiting a weight decay method to help drastically reducing the learning rate near optima to reach the dominant local minimum. We demonstrate the superiority of our proposed work by surpassing the first rank of three widely used Facial Expression Recognition datasets with 93.30% on FER-2013, and 16% improvement compare to the first rank after 10 years, reaching to 90.73% on RAF-DB, and 100% k-fold average accuracy for CK+ dataset, and shown to provide a top performance to that provided by other networks, which require much larger training datasets.

Keywords: computer vision, facial expression recognition, machine learning, algorithms, depp learning, neural networks

Procedia PDF Downloads 74

Authors: Asmaa Tahiri, Mouktafi Amine

Abstract:

A substantial number of traditional medical research has been put forward to managing and treating mental disorders. At the present time, to our best knowledge, it is believed that fundamental understanding of the underlying causes of the majority psychological disorders needs to be explored further to inform early diagnosis, managing symptoms and treatment. The emerging field of evolutionary psychology is a promising prospect to address the origin of mental disorders, potentially leading to more effective treatments. Schizophrenia as a topical mental disorder has been linked to the evolutionary adaptation of the human brain represented in the brain connectivity and asymmetry directly linked to humans higher brain cognition in contrast to other primates being our direct living representation of the structure and connectivity of our earliest common African ancestors. As proposed in the evolutionary psychology scientific literature the pathophysiology of schizophrenia is expressed and directly linked to altered connectivity between the Hippocampal Formation (HF) and Dorsolateral Prefrontal Cortex (DLPFC). This research paper presents the results of the use of tractography analysis using multiple open access Diffusion Weighted Imaging (DWI) datasets of healthy subjects, schizophrenia-affected subjects and primates to illustrate the relevance of the aforementioned brain regions connectivity and the underlying evolutionary changes in the human brain. Deterministic fiber tracking and streamline analysis were used to generate connectivity matrices from the DWI datasets overlaid to compute distances and highlight disconnectivity patterns in conjunction with other fiber tracking metrics; Fractional Anisotropy (FA), Mean Diffusivity (MD) and Radial Diffusivity (RD).

Keywords: tractography, evolutionary psychology, schizophrenia, brain connectivity

Procedia PDF Downloads 71

Authors: Danna Jia, Bin Li

Abstract:

Background: Atopic dermatitis (AD) is a chronic and refractory inflammatory skin disease characterized by relapsing eczematous and pruritic skin lesions. The global prevalence of AD ranges from 1~ 20%, and its incidence rates are increasing. It affects individuals from infancy to adulthood, significantly impacting their daily lives and social activities. Despite its major health burden, the precise mechanisms underlying AD remain unknown. Understanding the genetic differences associated with AD is crucial for advancing diagnosis and targeted treatment development. This study aims to identify candidate genes of AD by using bioinformatics analysis. Methods: We conducted a comprehensive analysis of four pooled transcriptomic datasets (GSE16161, GSE32924, GSE130588, and GSE120721) obtained from the Gene Expression Omnibus (GEO) database. Differential gene expression analysis was performed using the R statistical language. The differentially expressed genes (DEGs) between AD patients and normal individuals were functionally analyzed using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment. Furthermore, a protein-protein interaction (PPI) network was constructed to identify candidate genes. Results: Among the patient-level gene expression datasets, we identified 114 shared DEGs, consisting of 53 upregulated genes and 61 downregulated genes. Functional analysis using GO and KEGG revealed that the DEGs were mainly associated with the negative regulation of transcription from RNA polymerase II promoter, membrane-related functions, protein binding, and the Human papillomavirus infection pathway. Through the PPI network analysis, we identified eight core genes: CD44, STAT1, HMMR, AURKA, MKI67, and SMARCA4. Conclusion: This study elucidates key genes associated with AD, providing potential targets for diagnosis and treatment. The identified genes have the potential to contribute to the understanding and management of AD. The bioinformatics analysis conducted in this study offers new insights and directions for further research on AD. Future studies can focus on validating the functional roles of these genes and exploring their therapeutic potential in AD. While these findings will require further verification as achieved with experiments involving in vivo and in vitro models, these results provided some initial insights into dysfunctional inflammatory and immune responses associated with AD. Such information offers the potential to develop novel therapeutic targets for use in preventing and treating AD.

Keywords: atopic dermatitis, bioinformatics, biomarkers, genes

Procedia PDF Downloads 82

Authors: Jeremy Jenrette, Z. Y. C. Liu, Pranav Chimote, Edward Fox, Trevor Hastie, Francesco Ferretti

Abstract:

Suitable shark conservation depends on well-informed population assessments. Direct methods such as scientific surveys and fisheries monitoring are adequate for defining population statuses, but species-specific indices of abundance and distribution coming from these sources are rare for most shark species. We can rapidly fill these information gaps by boosting media-based remote monitoring efforts with machine learning and automation. We created a database of shark images by sourcing 24,546 images covering 219 species of sharks from the web application spark pulse and the social network Instagram. We used object detection to extract shark features and inflate this database to 53,345 images. We packaged object-detection and image classification models into a Shark Detector bundle. We developed the Shark Detector to recognize and classify sharks from videos and images using transfer learning and convolutional neural networks (CNNs). We applied these models to common data-generation approaches of sharks: boosting training datasets, processing baited remote camera footage and online videos, and data-mining Instagram. We examined the accuracy of each model and tested genus and species prediction correctness as a result of training data quantity. The Shark Detector located sharks in baited remote footage and YouTube videos with an average accuracy of 89\%, and classified located subjects to the species level with 69\% accuracy (n =\ eight species). The Shark Detector sorted heterogeneous datasets of images sourced from Instagram with 91\% accuracy and classified species with 70\% accuracy (n =\ 17 species). Data-mining Instagram can inflate training datasets and increase the Shark Detector’s accuracy as well as facilitate archiving of historical and novel shark observations. Base accuracy of genus prediction was 68\% across 25 genera. The average base accuracy of species prediction within each genus class was 85\%. The Shark Detector can classify 45 species. All data-generation methods were processed without manual interaction. As media-based remote monitoring strives to dominate methods for observing sharks in nature, we developed an open-source Shark Detector to facilitate common identification applications. Prediction accuracy of the software pipeline increases as more images are added to the training dataset. We provide public access to the software on our GitHub page.

Keywords: classification, data mining, Instagram, remote monitoring, sharks

Procedia PDF Downloads 121