Search results for: convolutional coding

Authors: Jingpeng Gao, Xinrui Mao, Zhibin Deng

Abstract:

Aiming at the low recognition rate on the composite signal modulation in low signal to noise ratio (SNR), this paper proposes a modulation recognition algorithm based on ICanny-CNN. Firstly, the radar signal is transformed into the time-frequency image by Choi-Williams Distribution (CWD). Secondly, we propose an image processing algorithm using the Guided Filter and the threshold selection method, which is combined with the hole filling and the mask operation. Finally, the shallow convolutional neural network (CNN) is combined with the idea of the depth-wise convolution (Dw Conv) and the point-wise convolution (Pw Conv). The proposed CNN is designed to complete image classification and realize modulation recognition of radar signal. The simulation results show that the proposed algorithm can reach 90.83% at 0dB and 71.52% at -8dB. Therefore, the proposed algorithm has a good classification and anti-noise performance in radar signal modulation recognition and other fields.

Keywords: modulation recognition, image processing, composite signal, improved Canny algorithm

Procedia PDF Downloads 191

Authors: Evrim Genc Kumtepe, Elif Toprak, Aylin Ozturk, Gamze Tuna, Hakan Kilinc, Irem Aydin Menderis

Abstract:

Support services are very significant elements for all educational institutions in general; however, for distance learners, these services are more essential than traditional (face-to-face) counterparts. One of the most important reasons for this is that learners and instructors do not share the same physical environment and that distance learning settings generally require intrapersonal interactions rather than interpersonal ones. Some learners in distance learning programs feel isolated. Furthermore, some fail to feel a sense of belonging to the institution because of lack of self-management skills, lack of motivation levels, and the need of being socialized, so that they are more likely to fail or drop out of an online class. In order to overcome all these problems, support services have emerged as a critical element for an effective and sustainable distance education system. Within the context of distance education support services, it is natural to include technology-based and web-based services and also the related materials. Moreover, institutions in education sector are expected to use information and communication technologies effectively in order to be successful in educational activities and programs. In terms of the sustainability of the system, an institution should provide distance education services through ICT enabled processes to support all stakeholders in the system, particularly distance learners. In this study, it is envisaged to develop a model based on the current support services literature in the field of open and distance learning and the applications of the distance higher education institutions. Specifically, content analysis technique is used to evaluate the existing literature in the distance education support services, the information published on websites, and applications of distance higher education institutions across the world. A total of 60 institutions met the inclusion criteria which are language option (English) and availability of materials in the websites. The six field experts contributed to brainstorming process to develop and extract codes for the coding scheme. During the coding process, these preset and emergent codes are used to conduct analyses. Two coders independently reviewed and coded each assigned website to ensure that all coders are interpreting the data the same way and to establish inter-coder reliability. Once each web page is included in descriptive and relational analysis, a model of support services is developed by examining the generated codes and themes. It is believed that such a model would serve as a quality guide for future institutions, as well as the current ones.

Keywords: support services, open education, distance learning, support model

Procedia PDF Downloads 197

Authors: Hrishi Rakshit, Pooneh Bagheri Zadeh

Abstract:

This paper investigates the performance of drone classification methods using conventional DCNN with different hyperparameters, when additional drone audio data is embedded in the dataset for training and further classification. In this paper, first a custom dataset is created using different images of drones from University of South California (USC) datasets and Leeds Beckett university datasets with embedded drone audio signal. The three well-known DCNN architectures namely, Resnet50, Darknet53 and Shufflenet are employed over the created dataset tuning their hyperparameters such as, learning rates, maximum epochs, Mini Batch size with different optimizers. Precision-Recall curves and F1 Scores-Threshold curves are used to evaluate the performance of the named classification algorithms. Experimental results show that Resnet50 has the highest efficiency compared to other DCNN methods.

Keywords: drone classifications, deep convolutional neural network, hyperparameters, drone audio signal

Procedia PDF Downloads 104

Authors: Donya Ashtiani Haghighi, Amirali Baniasadi

Abstract:

Coronavirus (COVID-19) disease has spread abruptly all over the world since the end of 2019. Computed tomography (CT) scans and X-ray images are used to detect this disease. Different Deep Neural Network (DNN)-based diagnosis solutions have been developed, mainly based on Convolutional Neural Networks (CNNs), to accelerate the identification of COVID-19 cases. However, CNNs lose important information in intermediate layers and require large datasets. In this paper, Capsule Network (CapsNet) is used. Capsule Network performs better than CNNs for small datasets. Accuracy of 0.9885, f1-score of 0.9883, precision of 0.9859, recall of 0.9908, and Area Under the Curve (AUC) of 0.9948 are achieved on the Capsule-based framework with hyperparameter tuning. Moreover, different dropout rates are investigated to decrease overfitting. Accordingly, a dropout rate of 0.1 shows the best results. Finally, we remove one convolution layer and decrease the number of trainable parameters to 146,752, which is a promising result.

Keywords: capsule network, dropout, hyperparameter tuning, classification

Procedia PDF Downloads 77

Authors: Hamida Qumber Ali, Waseem Muhammad Arain, Shama Siddiqui, Sayeed Ghani

Abstract:

We present an implementing of RMAC in TinyOS 1.x. RMAC is a cross layer and Duty-cycle MAC protocols that was proposed to provide energy efficient transmission services for wireless sensor networks. The protocol has a unique and efficient packet transmission scheduling mechanism that enables it to overcome delivery latency and overcome traffic congestion. Design details and implementation challenges are divulged. Experiments are conducted to show the correctness of our implementation with numerous assumptions. Simulations are performed to compare the performance of RMAC and SMAC. Our results show that RMAC outperforms SMAC in energy efficiency and delay.

Keywords: MAC protocol, performance, RMAC, wireless sensor networks

Procedia PDF Downloads 325

Authors: Rose Antony, Vivekanand B. Khanapuri, Karuna Jain

Abstract:

Online groceries have transformed the way the supply chains are managed. These are facing numerous challenges in terms of product wastages, low margins, long breakeven to achieve and low market penetration to mention a few. The e-grocery chains need to overcome these challenges in order to survive the competition. The purpose of this paper is to carry out a structural analysis of the enablers in e-grocery chains by applying Interpretive Structural Modeling (ISM) and MICMAC analysis in the Indian context. The research design is descriptive-explanatory in nature. The enablers have been identified from the literature and through semi-structured interviews conducted among the managers having relevant experience in e-grocery supply chains. The experts have been contacted through professional/social networks by adopting a purposive snowball sampling technique. The interviews have been transcribed, and manual coding is carried using open and axial coding method. The key enablers are categorized into themes, and the contextual relationship between these and the performance measures is sought from the Industry veterans. Using ISM, the hierarchical model of the enablers is developed and MICMAC analysis identifies the driver and dependence powers. Based on the driver-dependence power the enablers are categorized into four clusters namely independent, autonomous, dependent and linkage. The analysis found that information technology (IT) and manpower training acts as key enablers towards reducing the lead time and enhancing the online service quality. Many of the enablers fall under the linkage cluster viz., frequent software updating, branding, the number of delivery boys, order processing, benchmarking, product freshness and customized applications for different stakeholders, depicting these as critical in online food/grocery supply chains. Considering the perishability nature of the product being handled, the impact of the enablers on the product quality is also identified. Hence, study aids as a tool to identify and prioritize the vital enablers in the e-grocery supply chain. The work is perhaps unique, which identifies the complex relationships among the supply chain enablers in fresh food for e-groceries and linking them to the performance measures. It contributes to the knowledge of supply chain management in general and e-retailing in particular. The approach focus on the fresh food supply chains in the Indian context and hence will be applicable in developing economies context, where supply chains are evolving.

Keywords: interpretive structural modelling (ISM), India, online grocery, retail operations, supply chain management

Procedia PDF Downloads 204

Authors: Eshta Ranyal, Kamal Jain, Vikrant Ranyal

Abstract:

Potholes are a severe threat to road safety and a major contributing factor towards road distress. In the Indian context, they are a major road hazard. Timely detection of potholes and subsequent repair can prevent the roads from deteriorating. To facilitate the roadway authorities in the timely detection and repair of potholes, we propose a pothole detection methodology using convolutional neural networks. The YOLOv3 model is used as it is fast and accurate in comparison to other state-of-the-art models. You only look once v3 (YOLOv3) is a state-of-the-art, real-time object detection system that features multi-scale detection. A mean average precision(mAP) of 73% was obtained on a training dataset of 200 images. The dataset was then increased to 500 images, resulting in an increase in mAP. We further calculated the depth of the potholes using stereoscopic vision by reconstruction of 3D potholes. This enables calculating pothole volume, its extent, which can then be used to evaluate the pothole severity as low, moderate, high.

Keywords: CNN, pothole detection, pothole severity, YOLO, stereovision

Procedia PDF Downloads 136

Authors: Sergey Ermolin, Olga Ermolin

Abstract:

A Deep Convolution Neural Network with Triplet Loss is used to identify duplicate images in real-estate advertisements in the presence of image artifacts such as watermarking, cropping, hue/brightness adjustment, and others. The effects of batch normalization, spatial dropout, and various convergence methodologies on the resulting detection accuracy are discussed. For comparative Return-on-Investment study (per industry request), end-2-end performance is benchmarked on both Nvidia Titan GPUs and Intel’s Xeon CPUs. A new real-estate dataset from San Francisco Bay Area is used for this work. Sufficient duplicate detection accuracy is achieved to supplement other database-grounded methods of duplicate removal. The implemented method is used in a Proof-of-Concept project in the real-estate industry.

Keywords: visual recognition, convolutional neural networks, triplet loss, spatial batch normalization with dropout, duplicate removal, advertisement technologies, performance benchmarking

Procedia PDF Downloads 338

Authors: Zohar Manber, Ran Elkon

Abstract:

Accumulation of somatic mutations (SMs) in the genome is a major driving force of cancer development. Most SMs in the tumor's genome are functionally neutral; however, some cause damage to critical processes and provide the tumor with a selective growth advantage (termed cancer driver mutations). Current research on functional significance of SMs is mainly focused on finding alterations in protein coding sequences. However, the exome comprises only 3% of the human genome, and thus, SMs in the noncoding genome significantly outnumber those that map to protein-coding regions. Although our understanding of noncoding driver SMs is very rudimentary, it is likely that disruption of regulatory elements in the genome is an important, yet largely underexplored mechanism by which somatic mutations contribute to cancer development. The expression of most human genes is controlled by multiple enhancers, and therefore, it is conceivable that regulatory SMs are distributed across different enhancers of the same target gene. Yet, to date, most statistical searches for regulatory SMs have considered each regulatory element individually, which may reduce statistical power. The first challenge in considering the cumulative activity of all the enhancers of a gene as a single unit is to map enhancers to their target promoters. Such mapping defines for each gene its set of regulating enhancers (termed "set of regulatory elements" (SRE)). Considering multiple enhancers of each gene as one unit holds great promise for enhancing the identification of driver regulatory SMs. However, the success of this approach is greatly dependent on the availability of comprehensive and accurate enhancer-promoter (E-P) maps. To date, the discovery of driver regulatory SMs has been hindered by insufficient sample sizes and statistical analyses that often considered each regulatory element separately. In this study, we analyzed more than 2,500 whole-genome sequence (WGS) samples provided by The Cancer Genome Atlas (TCGA) and The International Cancer Genome Consortium (ICGC) in order to identify such driver regulatory SMs. Our analyses took into account the combinatorial aspect of gene regulation by considering all the enhancers that control the same target gene as one unit, based on E-P maps from three genomics resources. The identification of candidate driver noncoding SMs is based on their recurrence. We searched for SREs of genes that are "hotspots" for SMs (that is, they accumulate SMs at a significantly elevated rate). To test the statistical significance of recurrence of SMs within a gene's SRE, we used both global and local background mutation rates. Using this approach, we detected - in seven different cancer types - numerous "hotspots" for SMs. To support the functional significance of these recurrent noncoding SMs, we further examined their association with the expression level of their target gene (using gene expression data provided by the ICGC and TCGA for samples that were also analyzed by WGS).

Keywords: cancer genomics, enhancers, noncoding genome, regulatory elements

Procedia PDF Downloads 104

Authors: Cecile Garry, Katia Rovira, Julie Brisson

Abstract:

Autism Spectrum Disorder (ASD) is defined by a qualitative and quantitative impairment of social interaction. Indeed early intervention programs, such as the Early Start Denver Model (ESDM), aimed at encouraging the development of social skills. In classroom, the children need to be socially engaged to learn. Early intervention programs can thus be implemented in kindergarten schools. In these schools, ASD children have more opportunities to interact with their peers or adults than in elementary schools. However, the preschool children with ASD are less socially engaged than their typically developing peers in the classroom. They initiate, respond and maintain less the social interactions. In addition, they produce more responses than initiations. When they interact, the non verbal communication is more used than verbal or symbolic communication forms and they are more engaged with adults than with peers. Nevertheless, communicative patterns may vary according to the clinical profiles of ASD children. Indeed, the ASD children with better cognitive skills interact more with their peers and use more symbolic communication than the ASD children with a low cognitive level. ASD children with the less severe symptoms use more the verbal communication than ASD children with the more severe symptoms. Small groups and structured activities encourage coordinated joint engagement episodes in ASD children. Our goal is to evaluate ASD children’s social engagement development in class, with their peers or adults, during dyadic or group activities. Participants were 19 preschool children with ASD aged from 3 to 6 years old that benefited of an early intervention in special kindergarten schools. Severity of ASD symptoms was measured with the CARS at the beginning of the follow-up. Classroom situations of interaction were recorded during 10 minutes (5 minutes of dyadic interaction and 5 minutes of a group activity), every 2 months, during 10 months. Social engagement behaviors of children, including initiations, responses and imitation, directed to a peer or an adult, were then coded. The Observer software (Noldus) that allows to annotate behaviors was the coding system used. A double coding was conducted and revealed a good inter judges fidelity. Results show that ASD children were more often and longer socially engaged in dyadic than in groups situations. They were also more engaged with adults than with peers. Children with the less severe symptoms of ASD were more socially engaged in groups situations than children with the more severe symptoms of ASD. Then, ASD children with the less severe symptoms of ASD were more engaged with their peers than ASD children with the more severe symptoms of ASD. However, the engagement frequency increased during the 10 month of follow-up but only for ASD children with the more severe symptoms at the beginning. To conclude, these results highlighted the necessity of individualizing early intervention programs according to the clinical profile of the child.

Keywords: autism spectrum disorder, preschool children, developmental psychology, early interventions, social interactions

Procedia PDF Downloads 159

Authors: Jia Li, Xing Wei, Yuchen Hong, Yang Lu

Abstract:

Detecting vehicle behavior has always been the focus of intelligent transportation, but with the explosive growth of the number of vehicles and the complexity of the road environment, the vehicle behavior videos captured by traditional surveillance have been unable to satisfy the study of vehicle behavior. The traditional method of manually labeling vehicle behavior is too time-consuming and labor-intensive, but the existing object detection and tracking algorithms have poor practicability and low behavioral location detection rate. This paper proposes a vehicle behavior detection algorithm based on the dual-stream convolution network and the multi-dimensional video dynamic detection network. In the videos, the straight-line behavior of the vehicle will default to the background behavior. The Changing lanes, turning and turning around are set as target behaviors. The purpose of this model is to automatically mark the target behavior of the vehicle from the untrimmed videos. First, the target behavior proposals in the long video are extracted through the dual-stream convolution network. The model uses a dual-stream convolutional network to generate a one-dimensional action score waveform, and then extract segments with scores above a given threshold M into preliminary vehicle behavior proposals. Second, the preliminary proposals are pruned and identified using the multi-dimensional video dynamic detection network. Referring to the hierarchical reinforcement learning, the multi-dimensional network includes a Timer module and a Spacer module, where the Timer module mines time information in the video stream and the Spacer module extracts spatial information in the video frame. The Timer and Spacer module are implemented by Long Short-Term Memory (LSTM) and start from an all-zero hidden state. The Timer module uses the Transformer mechanism to extract timing information from the video stream and extract features by linear mapping and other methods. Finally, the model fuses time information and spatial information and obtains the location and category of the behavior through the softmax layer. This paper uses recall and precision to measure the performance of the model. Extensive experiments show that based on the dataset of this paper, the proposed model has obvious advantages compared with the existing state-of-the-art behavior detection algorithms. When the Time Intersection over Union (TIoU) threshold is 0.5, the Average-Precision (MP) reaches 36.3% (the MP of baselines is 21.5%). In summary, this paper proposes a vehicle behavior detection model based on multi-dimensional dynamic detection network. This paper introduces spatial information and temporal information to extract vehicle behaviors in long videos. Experiments show that the proposed algorithm is advanced and accurate in-vehicle timing behavior detection. In the future, the focus will be on simultaneously detecting the timing behavior of multiple vehicles in complex traffic scenes (such as a busy street) while ensuring accuracy.

Keywords: vehicle behavior detection, convolutional neural network, long short-term memory, deep learning

Procedia PDF Downloads 130

Authors: Saba Rahimi, Ozan Oktay, Javier Alvarez-Valle, Sujeeth Bharadwaj

Abstract:

Successful application of deep learning in medical image analysis necessitates unprecedented amounts of labeled training data. Unlike conventional 2D applications, radiological images can be three-dimensional (e.g., CT, MRI), consisting of many instances within each image. The problem is exacerbated when expert annotations are required for effective pixel-wise labeling, which incurs exorbitant labeling effort and cost. Active learning is an established research domain that aims to reduce labeling workload by prioritizing a subset of informative unlabeled examples to annotate. Our contribution is a cost-effective approach for U-Net 3D models that uses Monte Carlo sampling to analyze pixel-wise uncertainty. Experiments on the AAPM 2017 lung CT segmentation challenge dataset show that our proposed framework can achieve promising segmentation results by using only 42% of the training data.

Keywords: image segmentation, active learning, convolutional neural network, 3D U-Net

Procedia PDF Downloads 155

Authors: Desta Kebede Ayana, Lies Sercu, Demelash Mengistu

Abstract:

The study is set in Ethiopia, a sub-Saharan multilingual, multiethnic African country, which has seen a significant increase in the number of universities in recent years. The aim of this growth is to provide access to education for all cultural and linguistic groups across the country. However, there are challenges in promoting intercultural competence among students in this diverse context. The aim of the study is to investigate the interculturalization of Ethiopian Higher Education Institutions as perceived by university lecturers and administrators. In particular, the study aims to determine the level of support for this educational innovation and gather suggestions for its implementation and institutionalization. The researchers employed semi-structured interviews with administrators and lecturers from two large Ethiopian universities to gather data. Thematic analysis was utilized for coding and analyzing the interview data, with the assistance of the NVIVO software. The findings obtained from the grounded analysis of the interview data reveal that while there are opportunities for interculturalization in the curriculum and campus life, support for educational innovation remains low. Administrators and lecturers also emphasize the government's responsibility to prioritize interculturalization over other educational innovation goals. The study contributes to the existing literature by examining an under-researched population in an under-researched context. Additionally, the study explores whether Western perspectives of intercultural competence align with the African context, adding to the theoretical understanding of intercultural education. The data for this study was collected through semi-structured interviews conducted with administrators and lecturers from two large Ethiopian universities. The interviews allowed for an in-depth exploration of the participants' views on interculturalization in higher education. Thematic analysis was applied to the interview data, allowing for the identification and organization of recurring themes and patterns. The analysis was conducted using the NVIVO software, which aided in coding and analyzing the data. The study addresses the extent to which administrators and lecturers support the interculturalization of Ethiopian Higher Education Institutions. It also explores their suggestions for implementing and institutionalizing intercultural education, as well as their perspectives on the current level of institutionalization. The study highlights the challenges in interculturalizing Ethiopian universities and emphasizes the need for greater support and prioritization of intercultural education. It also underscores the importance of considering the African context when conceptualizing intercultural competence. This research contributes to the understanding of intercultural education in diverse contexts and provides valuable insights for policymakers and educational institutions aiming to promote intercultural competence in higher education settings.

Keywords: administrators, educational change, Ethiopia, intercultural competence, lecturers

Procedia PDF Downloads 97

Authors: Li Kewen, Su Zhaoxin, Wang Xingmou, Zhu Jian Bing

Abstract:

Conventional reservoir prediction methods ar not sufficient to explore the implicit relation between seismic attributes, and thus data utilization is low. In order to improve the predictive classification accuracy of reservoir lithology, this paper proposes a deep learning lithology prediction method based on ResNet (Residual Neural Network) and SENet (Squeeze-and-Excitation Neural Network). The neural network model is built and trained by using seismic attribute data and lithology data of Shengli oilfield, and the nonlinear mapping relationship between seismic attribute and lithology marker is established. The experimental results show that this method can significantly improve the classification effect of reservoir lithology, and the classification accuracy is close to 70%. This study can effectively predict the lithology of undrilled area and provide support for exploration and development.

Keywords: convolutional neural network, lithology, prediction of reservoir, seismic attributes

Procedia PDF Downloads 177

Authors: A. Van Staden, A. Tolmie, E. Vorster

Abstract:

Research confirms the multifaceted nature of spelling development and underscores the importance of both cognitive and linguistic skills that affect sound spelling development such as working and long-term memory, phonological and orthographic awareness, mental orthographic images, semantic knowledge and morphological awareness. This has clear implications for many South African English second language spellers (L2) who attempt to become proficient spellers. Since English has an opaque orthography, with irregular spelling patterns and insufficient sound/grapheme correspondences, L2 spellers can neither rely, nor draw on the phonological awareness skills of their first language (for example Sesotho and many other African languages), to assist them to spell the majority of English words. Epistemologically, this research is informed by social constructivism. In addition the researchers also hypothesized that the principles of the Overlapping Waves Theory was an appropriate lens through which to investigate whether L2 spellers could significantly improve their spelling skills via the implementation of an alternative route to spelling development, namely the orthographic route, and more specifically via the application of visual imagery. Post-test results confirmed the results of previous research that argues for the interactive nature of different cognitive and linguistic systems such as working memory and its subsystems and long-term memory, as learners were systematically guided to store visual orthographic images of words in their long-term lexicons. Moreover, the results have shown that L2 spellers in the experimental group (n = 9) significantly outperformed L2 spellers (n = 9) in the control group whose intervention involved phonological awareness (and coding) including the teaching of spelling rules. Consequently, L2 learners in the experimental group significantly improved in all the post-test measures included in this investigation, namely the four sub-tests of short-term memory; as well as two spelling measures (i.e. diagnostic and standardized measures). Against this background, the findings of this study look promising and have shown that, within a social-constructivist learning environment, learners can be systematically guided to apply higher-order thinking processes such as visual imagery to successfully store and retrieve mental images of spelling words from their output lexicons. Moreover, results from the present study could play an important role in directing research into this under-researched aspect of L2 literacy development within the South African education context.

Keywords: English second language spellers, phonological and orthographic coding, social constructivism, visual imagery as spelling strategy

Procedia PDF Downloads 359

Authors: Ji Su Kim, Bo Ram Choi, Ju Yeon Cho, Hyukjin Lee

Abstract:

Epidermal growth factor receptor (EGFR) 19 deletion mutation gene is over-expressed in carcinoma patient. EGFR 19 deletion mutation is known as typical biomarker of non-small cell lung cancer (NSCLC), which one section in the coding exon 19 of EGFR is deleted. Therefore, there have been many attempts over the years to detect EGFR 19 deletion mutation for replacing conventional diagnostic method such as PCR and tissue biopsy. We developed a simple and facile detection platform based on Rolling Circle Amplification (RCA), which provides highly amplified products in isothermal amplification of the ligated DNA template. Limit of detection (~50 nM) and a faster detection time (~30 min) could be achieved by introducing RCA.

Keywords: EGFR19, cancer, diagnosis, rolling circle amplification (RCA), hydrogel

Procedia PDF Downloads 255

Authors: Akeel A. Shah, Tong Zhang

Abstract:

Computational approaches to learning the properties of materials are commonplace, motivated by the need to screen or design materials for a given application, e.g., semiconductors and energy storage. Experimental approaches can be both time consuming and costly. Unfortunately, computational approaches such as ab-initio electronic structure calculations and classical or ab-initio molecular dynamics are themselves can be too slow for the rapid evaluation of materials, often involving thousands to hundreds of thousands of candidates. Machine learning assisted approaches have been developed to overcome the time limitations of purely physics-based approaches. These approaches, on the other hand, require large volumes of data for training (hundreds of thousands on many standard data sets such as QM7b). This means that they are limited by how quickly such a large data set of physics-based simulations can be established. At high fidelity, such as configuration interaction, composite methods such as G4, and coupled cluster theory, gathering such a large data set can become infeasible, which can compromise the accuracy of the predictions - many applications require high accuracy, for example band structures and energy levels in semiconductor materials and the energetics of charge transfer in energy storage materials. In order to circumvent this problem, multi-fidelity approaches can be adopted, for example the Δ-ML method, which learns a high-fidelity output from a low-fidelity result such as Hartree-Fock or density functional theory (DFT). The general strategy is to learn a map between the low and high fidelity outputs, so that the high-fidelity output is obtained a simple sum of the physics-based low-fidelity and correction, Although this requires a low-fidelity calculation, it typically requires far fewer high-fidelity results to learn the correction map, and furthermore, the low-fidelity result, such as Hartree-Fock or semi-empirical ZINDO, is typically quick to obtain, For high-fidelity outputs the result can be an order of magnitude or more in speed up. In this work, a new multi-fidelity approach is developed, based on a graph convolutional network (GCN) combined with semi-supervised learning. The GCN allows for the material or molecule to be represented as a graph, which is known to improve accuracy, for example SchNet and MEGNET. The graph incorporates information regarding the numbers of, types and properties of atoms; the types of bonds; and bond angles. They key to the accuracy in multi-fidelity methods, however, is the incorporation of low-fidelity output to learn the high-fidelity equivalent, in this case by learning their difference. Semi-supervised learning is employed to allow for different numbers of low and high-fidelity training points, by using an additional GCN-based low-fidelity map to predict high fidelity outputs. It is shown on 4 different data sets that a significant (at least one order of magnitude) increase in accuracy is obtained, using one to two orders of magnitude fewer low and high fidelity training points. One of the data sets is developed in this work, pertaining to 1000 simulations of quinone molecules (up to 24 atoms) at 5 different levels of fidelity, furnishing the energy, dipole moment and HOMO/LUMO.

Keywords: .materials screening, computational materials, machine learning, multi-fidelity, graph convolutional network, semi-supervised learning

Procedia PDF Downloads 40

Authors: Pingping Lin, Xudong Luo, Yifan Fan

Abstract:

Sentiment analysis is a very active research topic. Every day, Facebook, Twitter, Weibo, and other social media, as well as significant e-commerce websites, generate a massive amount of comments, which can be used to analyse peoples opinions or emotions. The existing methods for sentiment analysis are based mainly on sentiment dictionaries, machine learning, and deep learning. The first two kinds of methods rely on heavily sentiment dictionaries or large amounts of labelled data. The third one overcomes these two problems. So, in this paper, we focus on the third one. Specifically, we survey various sentiment analysis methods based on convolutional neural network, recurrent neural network, long short-term memory, deep neural network, deep belief network, and memory network. We compare their futures, advantages, and disadvantages. Also, we point out the main problems of these methods, which may be worthy of careful studies in the future. Finally, we also examine the application of deep learning in multimodal sentiment analysis and aspect-level sentiment analysis.

Keywords: document analysis, deep learning, multimodal sentiment analysis, natural language processing

Procedia PDF Downloads 164

Authors: Justin Reverdi, Sixin Zhang, Serge Gratton, Said Aoues, Thomas Pellegrini

Abstract:

In Vapor Cycle Systems, the flow sensor plays a key role in different monitoring and control purposes. However, physical sensors can be expensive, inaccurate, heavy, cumbersome, or highly sensitive to vibrations, which is especially problematic when embedded into an aircraft. The conception of a virtual sensor based on other standard sensors is a good alternative. In this paper, a data-driven model using a Convolutional Neural Network is proposed to estimate the flow of the compressor. To fit the model to our dataset, we tested different loss functions. We show in our application that a Dynamic Time Warping based loss function called DILATE leads to better dynamical performance than the vanilla mean squared error (MSE) loss function. DILATE allows choosing a trade-off between static and dynamic performance.

Keywords: deep learning, dynamic time warping, vapor cycle system, virtual sensor

Procedia PDF Downloads 146

Authors: Filip Majetić, Dražen Šimleša, Jelena Puđak, Anita Bušljeta Tonković, Svitlana Pinchuk

Abstract:

This paper focuses on the self-sustainability aspect of social entrepreneurship (SE). We define SE as a form of entrepreneurship that is social/ecological mission oriented. It means SE organizations start and run businesses and use them to accomplish their social/ecological missions i.e. to solve social/ecological problems or fulfill social/ecological needs. Self-sustainability is defined as the capability of an SE organization to operate by relying on the money earned through trading its products in the free market. For various reasons, the achievement of self-sustainability represents a fundamental (business) challenge for many SE organizations. Those that are not able to operate using the money made through commercial activities, in order to remain active, rely on alternative, non-commercial streams of income such as grants, donations, and public subsidies. Starting from this widespread (business) challenge, we are interested in exploring elements that (could) influence the self-sustainability in SE organizations. Therefore, the research goal is to empirically investigate some of the self-sustainability factors of two notable SE organizations from different socio-economic contexts. A qualitative research, using the multiple case study approach, was conducted. ACT Group Čakovec (ACT) from Croatia was selected for the first case because it represents one of the leading and most self-sustainable SE organization in the region (in 2015 55% of the organization’s budget came from commercial activities); Friskis&Svettis Stockholm (F&S) from Sweden was selected for the second case because it is a rare example of completely self-sustainable SE organization in Europe (100% of the organization’s budget comes from commercial activities). The data collection primarily consists of conducting in-depth interviews. Additionally, the content of some of the organizations' official materials are analyzed (e.g. business reports, marketing materials). The interviewees are selected purposively and include: six highly ranked F&S members who represent five different levels in the hierarchy of their organization; five highly ranked ACT members who represent three different levels in the hierarchy of the organization. All of the interviews contain five themes: a) social values of the organization, b) organization of work, c) non-commercial income sources, d) marketing/collaborations, and e) familiarity with the industry characteristics and trends. The gathered data is thematically analyzed through the coding process for which Atlas.ti software for qualitative data analysis is used. For the purpose of creating thematic categories (codes), the open coding is used. The research results intend to provide new theoretical insights on factors of SE self-sustainability and, preferably, encourage practical improvements in the field.

Keywords: Friskis&Svettis, self-sustainability factors, social entrepreneurship, Stockholm

Procedia PDF Downloads 218

Authors: Auwal Nata'ala

Abstract:

The Information and Communication Technology (ICT) has had impacts in almost every area human endeavor. From business, industries, banks to none profit organizations. ICT has simplified business process such as sorting, summarizing, coding, updating and generating a report in a real-time processing mode. However, the use of these ICT facilities such as computer and internet has also brought unintended consequences of criminal activities such as spamming, credit card frauds, ATM frauds, phishing, identity theft, denial of services and other related cyber crimes. This study sought to examined cyber-crime and its impact on the banking institution in Nigeria. It also examined the existing policy framework and assessed the success of the institutional countermeasures in combating cyber crime in the banking industry. This paper X-ray’s cyber crimes, policies issues and provides insight from a Nigeria perspective.

Keywords: cyber crimes, e-banking, policies, ICT

Procedia PDF Downloads 407

Authors: Hongmei Wang, Ziyun Xiang, Ying liu, Li Yu, Dongsheng Yue

Abstract:

Intro: The COVID-19 (Corona Virus Disease 2019) has greatly changed the global economic, political and financial ecology. The mutation of the coronavirus in the UK in December 2020 has brought new panic to the world. Deep learning was performed on Chest Computed tomography (CT) of COVID-19 and Influenza and describes their characteristics. The predominant features of COVID-19 pneumonia was ground-glass opacification, followed by consolidation. Lesion density: most lesions appear as ground-glass shadows, and some lesions coexist with solid lesions. Lesion distribution: the focus is mainly on the dorsal side of the periphery of the lung, with the lower lobe of the lungs as the focus, and it is often close to the pleura. Other features it has are grid-like shadows in ground glass lesions, thickening signs of diseased vessels, air bronchi signs and halo signs. The severe disease involves whole bilateral lungs, showing white lung signs, air bronchograms can be seen, and there can be a small amount of pleural effusion in the bilateral chest cavity. At the same time, this year's flu season could be near its peak after surging throughout the United States for months. Chest CT for Influenza infection is characterized by focal ground glass shadows in the lungs, with or without patchy consolidation, and bronchiole air bronchograms are visible in the concentration. There are patchy ground-glass shadows, consolidation, air bronchus signs, mosaic lung perfusion, etc. The lesions are mostly fused, which is prominent near the hilar and two lungs. Grid-like shadows and small patchy ground-glass shadows are visible. Deep neural networks have great potential in image analysis and diagnosis that traditional machine learning algorithms do not. Method: Aiming at the two major infectious diseases COVID-19 and influenza, which are currently circulating in the world, the chest CT of patients with two infectious diseases is classified and diagnosed using deep learning algorithms. The residual network is proposed to solve the problem of network degradation when there are too many hidden layers in a deep neural network (DNN). The proposed deep residual system (ResNet) is a milestone in the history of the Convolutional neural network (CNN) images, which solves the problem of difficult training of deep CNN models. Many visual tasks can get excellent results through fine-tuning ResNet. The pre-trained convolutional neural network ResNet is introduced as a feature extractor, eliminating the need to design complex models and time-consuming training. Fastai is based on Pytorch, packaging best practices for in-depth learning strategies, and finding the best way to handle diagnoses issues. Based on the one-cycle approach of the Fastai algorithm, the classification diagnosis of lung CT for two infectious diseases is realized, and a higher recognition rate is obtained. Results: A deep learning model was developed to efficiently identify the differences between COVID-19 and influenza using chest CT.

Keywords: COVID-19, Fastai, influenza, transfer network

Procedia PDF Downloads 142

Authors: Jutta Hämmerle-Uhl, Georg Penn, Gerhard Pötzelsberger, Andreas Uhl

Abstract:

Iris codes contain bits with different entropy. This work investigates different strategies to reduce the size of iris code templates with the aim of reducing storage requirements and computational demand in the matching process. Besides simple sub-sampling schemes, also a binary multi-resolution representation as used in the JBIG hierarchical coding mode is assessed. We find that iris code template size can be reduced significantly while maintaining recognition accuracy. Besides, we propose a two stage identification approach, using small-sized iris code templates in a pre-selection satge, and full resolution templates for final identification, which shows promising recognition behaviour.

Keywords: iris recognition, compact iris code, fast matching, best bits, pre-selection identification, two-stage identification

Procedia PDF Downloads 440

Authors: Nilubon Kurubanjerdjit, Nattakarn Iam-On, Ka-Lok Ng

Abstract:

MicroRNAs are small non-coding RNA found in many different species. They play crucial roles in cancer such as biological processes of apoptosis and proliferation. The identification of microRNA-target genes can be an essential first step towards to reveal the role of microRNA in various cancer types. In this paper, we predict miRNA-target genes for lung cancer by integrating prediction scores from miRanda and PITA algorithms used as a feature vector of miRNA-target interaction. Then, machine-learning algorithms were implemented for making a final prediction. The approach developed in this study should be of value for future studies into understanding the role of miRNAs in molecular mechanisms enabling lung cancer formation.

Keywords: microRNA, miRNAs, lung cancer, machine learning, Naïve Bayes, SVM

Procedia PDF Downloads 399

Authors: Bandhan Dey, Muhsina Bintoon Yiasha, Gulam Sulaman Choudhury

Abstract:

Chest disease is one of the most problematic ailments in our regular life. There are many known chest diseases out there. Diagnosing them correctly plays a vital role in the process of treatment. There are many methods available explicitly developed for different chest diseases. But the most common approach for diagnosing these diseases is through X-ray. In this paper, we proposed a multi-classification deep learning model for diagnosing COVID-19, lung cancer, pneumonia, tuberculosis, and atelectasis from chest X-rays. In the present work, we used the transfer learning method for better accuracy and fast training phase. The performance of three architectures is considered: InceptionV3, VGG-16, and VGG-19. We evaluated these deep learning architectures using public digital chest x-ray datasets with six classes (i.e., COVID-19, lung cancer, pneumonia, tuberculosis, atelectasis, and normal). The experiments are conducted on six-classification, and we found that VGG16 outperforms other proposed models with an accuracy of 95%.

Keywords: deep learning, image classification, X-ray images, Tensorflow, Keras, chest diseases, convolutional neural networks, multi-classification

Procedia PDF Downloads 92

Authors: McClain Thiel

Abstract:

Object detection and localization has historically required two or more sensors due to the loss of information from 3D to 2D space, however, most surveillance systems currently in use in the real world only have one sensor per location. Generally, this consists of a single low-resolution camera positioned above the area under observation (mall, jewelry store, traffic camera). This is not sufficient for robust 3D tracking for applications such as security or more recent relevance, contract tracing. This paper proposes a lightweight system for 3D person tracking that requires no additional hardware, based on compressed object detection convolutional-nets, facial landmark detection, and projective geometry. This approach involves classifying the target into a demographic category and then making assumptions about the relative locations of facial landmarks from the demographic information, and from there using simple projective geometry and known constants to find the target's location in 3D space. Preliminary testing, although severely lacking, suggests reasonable success in 3D tracking under ideal conditions.

Keywords: monocular distancing, computer vision, facial analysis, 3D localization

Procedia PDF Downloads 139

Authors: M. Hilani, B. Nassih

Abstract:

Phishing is one of the oldest and best known scams on the Internet. It can be defined as any type of telecommunications fraud that uses social engineering tricks to obtain confidential data from its victims. It’s a cybercrime aimed at stealing your sensitive information. Phishing is generally done via private email, so scammers impersonate large companies or other trusted entities to encourage victims to voluntarily provide information such as login credentials or, worse yet, credit card numbers. The COVID-19 theme is used by cybercriminals in multiple malicious campaigns like phishing. In this environment, messaging filtering solutions have become essential to protect devices that will now be used outside of the secure perimeter. Despite constantly updating methods to avoid these cyberattacks, the end result is currently insufficient. Many researchers are looking for optimal solutions to filter phishing emails, but we still need good results. In this work, we concentrated on solving the problem of detecting phishing emails using the different steps of NLP preprocessing, and we proposed and trained a model using one-dimensional CNN. Our study results show that our model obtained an accuracy of 99.99%, which demonstrates how well our model is working.

Keywords: phishing, e-mail, NLP preprocessing, CNN, e-mail filtering

Procedia PDF Downloads 126

Authors: Chad Goldsworthy, B. Rajeswari Matam

Abstract:

The use of deep learning for species identification in camera trap images has revolutionised our ability to study, conserve and monitor species in a highly efficient and unobtrusive manner, with state-of-the-art models achieving accuracies surpassing the accuracy of manual human classification. The high imbalance of camera trap datasets, however, results in poor accuracies for minority (rare or endangered) species due to their relative insignificance to the overall model accuracy. This paper investigates the use of Focal Loss, in comparison to the traditional Cross Entropy Loss function, to improve the identification of minority species in the “255 Bird Species” dataset from Kaggle. The results show that, although Focal Loss slightly decreased the accuracy of the majority species, it was able to increase the F1-score by 0.06 and improve the identification of the bottom two, five and ten (minority) species by 37.5%, 15.7% and 10.8%, respectively, as well as resulting in an improved overall accuracy of 2.96%.

Keywords: convolutional neural networks, data imbalance, deep learning, focal loss, species classification, wildlife conservation

Procedia PDF Downloads 191

Authors: Tushar K. Routh

Abstract:

If inputs are engineered in certain manners, they can influence deep neural networks’ (DNN) performances by facilitating misclassifications, a phenomenon well-known as adversarial attacks that question networks’ vulnerability. Recent studies have unfolded the relationship between vulnerability of such networks with their complexity. In this paper, the distinctive influence of additional convolutional layers at the decision boundaries of several DNN architectures was investigated. Here, to engineer inputs from widely known image datasets like MNIST, Fashion MNIST, and Cifar 10, we have exercised One Step Spectral Attack (OSSA) and Fast Gradient Method (FGM) techniques. The aftermaths of adding layers to the robustness of the architectures have been analyzed. For reasoning, separation width from linear class partitions and local geometry (curvature) near the decision boundary have been examined. The result reveals that model complexity has significant roles in adjusting relative distances from margins, as well as the local features of decision boundaries, which impact robustness.

Keywords: DNN robustness, decision boundary, local curvature, network complexity

Procedia PDF Downloads 75

Authors: Sham Kidane, Fitsum Gaim, Ibrahim Abdella, Sirak Asmerom, Yoel Ghebrihiwot, Simon Mulugeta, Natnael Ambassager

Abstract:

The lack of annotated datasets is a bottleneck to the progress of NLP in low-resourced languages. The work presented here consists of large-scale annotated datasets and models for the named entity recognition (NER) system for the Tigrinya language. Our manually constructed corpus comprises over 340K words tagged for NER, with over 118K of the tokens also having parts-of-speech (POS) tags, annotated with 12 distinct classes of entities, represented using several types of tagging schemes. We conducted extensive experiments covering convolutional neural networks and transformer models; the highest performance achieved is 88.8% weighted F1-score. These results are especially noteworthy given the unique challenges posed by Tigrinya’s distinct grammatical structure and complex word morphologies. The system can be an essential building block for the advancement of NLP systems in Tigrinya and other related low-resourced languages and serve as a bridge for cross-referencing against higher-resourced languages.

Keywords: Tigrinya NER corpus, TiBERT, TiRoBERTa, BiLSTM-CRF

Procedia PDF Downloads 130