Search results for: baselines

Authors: Gamal F.Attia

Abstract:

The baselines are the distances and lengths of the chords between projections of the positions of the laser stations on the reference ellipsoid. For the satellite geodesy, it is very important to determine the optimal length of orbital arc along which laser measurements are to be carried out. It is clear that for the dynamical methods long arcs (one month or more) are to be used. According to which more errors of modeling of different physical forces such as earth's gravitational field, air drag, solar radiation pressure, and others that may influence the accuracy of the estimation of the satellites position, at the same time the measured errors con be almost completely excluded and high stability in determination of relative coordinate system can be achieved. It is possible to diminish the influence of the errors of modeling by using short-arcs of the satellite orbit (several revolutions or days), but the station's coordinates estimated by different arcs con differ from each other by a larger quantity than statistical zero. Under the semidynamical ‘short arc’ method one or several passes of the satellite in one of simultaneous visibility from both ends of the chord is known and the estimated parameter in this case is the length of the chord. The comparison of the same baselines calculated with long and short arcs methods shows a good agreement and even speaks in favor of the last one. In this paper the Short Arc technique has been explained and 3 baselines have been determined using the ‘short arc’ method.

Keywords: baselines, short arc, dynamical, gravitational field

Procedia PDF Downloads 462

Authors: Xiaohan Bookman, Xiaoyan Zhu

Abstract:

The outputs of non-linear feed-forward neural network are positive, which could be treated as probability when they are normalized to one. If we take Entropy-Based Principle into consideration, the outputs for each sample could be represented as the distribution of this sample for different clusters. Entropy-Based Principle is the principle with which we could estimate the unknown distribution under some limited conditions. As this paper defines two processes in Feed-Forward Neural Network, our limited condition is the abstracted features of samples which are worked out in the abstraction process. And the final outputs are the probability distribution for different clusters in the clustering process. As Entropy-Based Principle is considered into the feed-forward neural network, a clustering method is born. We have conducted some experiments on six open UCI data sets, comparing with a few baselines and applied purity as the measurement. The results illustrate that our method outperforms all the other baselines that are most popular clustering methods.

Keywords: feed-forward neural network, clustering, max-entropy principle, probabilistic models

Procedia PDF Downloads 434

Authors: Lizhi Ma, Chengcheng Hu, Fuxian Wong

Abstract:

Partial reduce (P-Reduce) has set a state-of-the-art performance on distributed machine learning in the heterogeneous environment over the All-Reduce architecture. The dynamic P-Reduce based on the exponential moving average (EMA) approach predicts all the intermediate model parameters, which raises unreliability. It is noticed that the approximation trick leads the wrong way to obtaining model parameters in all the nodes. In this paper, SAP-Reduce is proposed, which is a variant of the All-Reduce distributed training model with staleness-aware dynamic P-Reduce. SAP-Reduce directly utilizes the EMA-like algorithm to generate the normalized weights. To demonstrate the effectiveness of the algorithm, the experiments are set based on a number of deep learning models, comparing the single-step training acceleration ratio and convergence time. It is found that SAP-Reduce simplifying dynamic P-Reduce outperforms the intermediate approximation one. The empirical results show SAP-Reduce is 1.3× −2.1× faster than existing baselines.

Keywords: collective communication, decentralized distributed training, machine learning, P-Reduce

Procedia PDF Downloads 31

Authors: Mohamed Ramdani, Hassen Abdellaoui, Abdenour Boudrassen

Abstract:

Mobile platform’s attitude estimation approaches mainly based on combined positioning techniques and developed algorithms; which aim to reach a fast and accurate solution. In this work, we describe the design and the implementation of an attitude determination (AD) process, using only measurements from GPS sensors. The major issue is based on smoothed GPS code data using Hatch filter and raw carrier-phase measurements integrated into attitude algorithm based on vectors measurement using least squares (LSQ) estimation method. GPS dataset from a static experiment is used to investigate the effectiveness of the presented approach and consequently to check the accuracy of the attitude estimation algorithm. Attitude results from GPS multi-antenna over short baselines are introduced and analyzed. The 3D accuracy of estimated attitude parameters using smoothed measurements is over 0.27°.

Keywords: attitude determination, GPS code data smoothing, hatch filter, carrier-phase measurements, least-squares attitude estimation

Procedia PDF Downloads 153

Authors: Nishtha Madaan, Prateek Chaudhury, Nishant Kumar, Srikanta Bedathur

Abstract:

In modern NLP applications, word embeddings are a crucial backbone that can be readily shared across a number of tasks. However, as the text distributions change and word semantics evolve over time, the downstream applications using the embeddings can suffer if the word representations do not conform to the data drift. Thus, maintaining word embeddings to be consistent with the underlying data distribution is a key problem. In this work, we tackle this problem and propose TransDrift, a transformer-based prediction model for word embeddings. Leveraging the flexibility of the transformer, our model accurately learns the dynamics of the embedding drift and predicts future embedding. In experiments, we compare with existing methods and show that our model makes significantly more accurate predictions of the word embedding than the baselines. Crucially, by applying the predicted embeddings as a backbone for downstream classification tasks, we show that our embeddings lead to superior performance compared to the previous methods.

Keywords: NLP applications, transformers, Word2vec, drift, word embeddings

Procedia PDF Downloads 88

Authors: Mohammad Lemar Zalmai, Osman Hurol Turkakin, Cemil Akcay, Ekrem Manisali

Abstract:

Most of the construction projects are exposed to time and cost overruns due to various factors and this is a major problem. As a solution to this, the Earned Value Management (EVM) method is considered. EVM is a powerful and well-known method used in monitoring and controlling the project. EVM is a technique that project managers use to track the performance of their project against project baselines. EVM gives an early indication that either project is delayed or not, and the project is either over budget or under budget at any particular day by tracking it. Thus, it helps to improve the management control system of a construction project, to detect and control the problems in potential risk areas and to suggest the importance and purpose of monitoring the construction work. This paper explains the main parameters of the EVM system involved in the calculation of time and cost for construction projects. In this study, the project management software Primavera P6 is used to deals with the project monitoring process of a seven-storeyed (G+6) faculty building whose construction is in progress at Istanbul, Turkey. A comparison between the planned progress of construction activities and actual progress is performed, and the analysis results are interpreted. This case study justifies the benefits of using EVM for project cash flow analysis and forecasting.

Keywords: earned value management (EVM), construction cost management, construction planning, primavera P6, project management, project scheduling

Procedia PDF Downloads 241

Authors: Pranjali Avinash Yadav-Deshmukh

Abstract:

Remarkable development of the Internet along with the new technological innovation, such as high-speed systems and affordable large storage space have led to a tremendous increase in the amount and accessibility to digital records. For any person, studying of all these data is tremendously time intensive, so there is a great need to access effective multi-document summarization (MDS) systems, which can successfully reduce details found in several records into a short, understandable summary or conclusion. For semantic representation of textual details in ontology area, as a theoretical design, our system provides a significant structure. The stability of using the ontology in fixing multi-document summarization problems in the sector of catastrophe control is finding its recommended design. Saliency ranking is usually allocated to each phrase and phrases are rated according to the ranking, then the top rated phrases are chosen as the conclusion. With regards to the conclusion quality, wide tests on a selection of media announcements are appropriate for “Jammu Kashmir Overflow in 2014” records. Ontology centered multi-document summarization methods using “NLP centered extraction” outshine other baselines. Our participation in recommended component is to implement the details removal methods (NLP) to enhance the results.

Keywords: disaster management, extraction technique, k-means, multi-document summarization, NLP, ontology, sentence extraction

Procedia PDF Downloads 385

Authors: Yunseok Noh, Chang-Uk Kwak, Sun-Joong Kim, Seong-Bae Park

Abstract:

Scripts are one of the basic text resources to understand broadcasting contents. Since broadcast media wields lots of influence over the public, tools for understanding broadcasting contents are more required. Topic modeling is the method to get the summary of the broadcasting contents from its scripts. Generally, scripts represent contents descriptively with directions and speeches. Scripts also provide scene segments that can be seen as semantic units. Therefore, a script can be topic modeled by treating a scene segment as a document. Because scripts consist of speeches mainly, however, relatively small co-occurrences among words in the scene segments are observed. This causes inevitably the bad quality of topics based on statistical learning method. To tackle this problem, we propose a method of learning with additional word co-occurrence information obtained using scene similarities. The main idea of improving topic quality is that the information that two or more texts are topically related can be useful to learn high quality of topics. In addition, by using high quality of topics, we can get information more accurate whether two texts are related or not. In this paper, we regard two scene segments are related if their topical similarity is high enough. We also consider that words are co-occurred if they are in topically related scene segments together. In the experiments, we showed the proposed method generates a higher quality of topics from Korean drama scripts than the baselines.

Keywords: broadcasting contents, scripts, text similarity, topic model

Procedia PDF Downloads 318

Authors: Jalil Qoulizadeh, Hasan Sadeghi

Abstract:

Brain waves are electrical wave patterns that are produced in the human brain. Knowing these waves and activating them can have a positive effect on brain function and ultimately create an ideal life. The brain has the ability to produce waves from 0.1 to above 65 Hz. (The Beta One device produces exactly these waves) This is because it is said that the waves produced by the Beta One device exactly match the waves produced by the brain. The function and method of this device is based on the magnetic stimulation of the brain. The technology used in the design and producƟon of this device works in a way to strengthen and improve the frequencies of brain waves with a pre-defined algorithm according to the type of requested function, so that the person can access the expected functions in life activities. to perform better. The effect of this field on neurons and their stimulation: In order to evaluate the effect of this field created by the device, on the neurons, the main tests are by conducting electroencephalography before and after stimulation and comparing these two baselines by qEEG or quantitative electroencephalography method using paired t-test in 39 subjects. It confirms the significant effect of this field on the change of electrical activity recorded after 30 minutes of stimulation in all subjects. The Beta One device is able to induce the appropriate pattern of the expected functions in a soft and effective way to the brain in a healthy and effective way (exactly in accordance with the harmony of brain waves), the process of brain activities first to a normal state and then to a powerful one. Production of inexpensive neuroscience equipment (compared to existing rTMS equipment) Magnetic brain stimulation for clinics - homes - factories and companies - professional sports clubs.

Keywords: stimulation, brain, waves, betaOne

Procedia PDF Downloads 80

Authors: Lizhi Ma, Dan Liu

Abstract:

Pipeline parallelism has been widely used to accelerate distributed deep learning to alleviate GPU memory bottlenecks and to ensure that models can be trained and deployed smoothly under limited graphics memory conditions. However, in highly heterogeneous distributed clusters, traditional model partitioning methods are not able to achieve load balancing. The overlap of communication and computation is also a big challenge. In this paper, HePipe is proposed, an efficient pipeline parallel training method for highly heterogeneous clusters. According to the characteristics of the neural network model pipeline training task, oriented to the 2-level heterogeneous cluster computing topology, a training method based on the 2-level stage division of neural network modeling and partitioning is designed to improve the parallelism. Additionally, a multi-forward 1F1B scheduling strategy is designed to accelerate the training time of each stage by executing the computation units in advance to maximize the overlap between the forward propagation communication and backward propagation computation. Finally, a dynamic recomputation strategy based on task memory requirement prediction is proposed to improve the fitness ratio of task and memory, which improves the throughput of the cluster and solves the memory shortfall problem caused by memory differences in heterogeneous clusters. The empirical results show that HePipe improves the training speed by 1.6×−2.2× over the existing asynchronous pipeline baselines.

Keywords: pipeline parallelism, heterogeneous cluster, model training, 2-level stage partitioning

Procedia PDF Downloads 17

Authors: Aschalew Cherie Workneh, K. S. Hari Prasad, Chandra Shekhar Prasad Ojha

Abstract:

Due to the increase in water scarcity, the crop water stress index (CWSI) is receiving significant attention these days, especially in arid and semiarid regions, for quantifying water stress and effective irrigation scheduling. Nowadays, machine learning techniques such as neural networks are being widely used to determine CWSI. In the present study, the performance of two artificial neural networks, namely, Self-Organizing Maps (SOM) and Feed Forward-Back Propagation Artificial Neural Networks (FF-BP-ANN), are compared while determining the CWSI of rice crop. Irrigation field experiments with varying degrees of irrigation were conducted at the irrigation field laboratory of the Indian Institute of Technology, Roorkee, during the growing season of the rice crop. The CWSI of rice was computed empirically by measuring key meteorological variables (relative humidity, air temperature, wind speed, and canopy temperature) and crop parameters (crop height and root depth). The empirically computed CWSI was compared with SOM and FF-BP-ANN predicted CWSI. The upper and lower CWSI baselines are computed using multiple regression analysis. The regression analysis showed that the lower CWSI baseline for rice is a function of crop height (h), air vapor pressure deficit (AVPD), and wind speed (u), whereas the upper CWSI baseline is a function of crop height (h) and wind speed (u). The performance of SOM and FF-BP-ANN were compared by computing Nash-Sutcliffe efficiency (NSE), index of agreement (d), root mean squared error (RMSE), and coefficient of correlation (R²). It is found that FF-BP-ANN performs better than SOM while predicting the CWSI of rice crops.

Keywords: artificial neural networks; crop water stress index; canopy temperature, prediction capability

Procedia PDF Downloads 116

Authors: Dejana Bouillet, Vlatka Domović

Abstract:

Both policymakers and researchers recognize that participating in early childhood education and care (ECEC) is useful for all children, especially for those who are exposed to the high risk of social exclusion. Social exclusion of children is understood as a multidimensional construct including economic, social, cultural, health, and other aspects of disadvantage and deprivation, which individually or combined can have an unfavorable effect on the current life and development of a child, as well as on the child’s development and on disadvantaged life chances in adult life. ECEC institutions should be able to promote educational approaches that portray developmental, cultural, language, and other diversity amongst children. However, little is known about the ways in which Croatian ECEC institutions recognize and respect the diversity of children and their families and how they respond to their educational needs. That is why this paper is dedicated to the analysis of the capacities of ECEC professionals to respond to the demands of educational needs of this very diverse group of children and their families. The results obtained in the frame of the project “Models of response to educational needs of children at risk of social exclusion in ECEC institutions,” funded by the Croatian Science Foundation, will be presented. The research methodology arises from explanations of educational processes and risks of social exclusion as a complex and heterogeneous phenomenon. The preliminary results of the qualitative data analysis of educational practices regarding capacities to identify and appropriately respond to the requirements of children at risk of social exclusion will be presented. The data have been collected by interviewing educational staff in 10 Croatian ECEC institutions (n = 10). The questions in the interviews were related to various aspects of inclusive institutional policy, culture, and practices. According to the analysis, it is possible to conclude that Croatian ECEC professionals are still faced with great challenges in the process of implementation of inclusive policies, culture, and practices. There are several baselines of this conclusion. The interviewed educational professionals are not familiar enough with the whole complexity and diversity of needs of children at risk of social exclusion, and the ECEC institutions do not have enough resources to provide all interventions that these children and their families need.

Keywords: children at risk of social exclusion, ECEC professionals, inclusive policies, culture and practices, quallitative analysis

Procedia PDF Downloads 113

Authors: Aschalew C. Workneh, K. S. Hari Prasad, C. S. P. Ojha

Abstract:

The crop water stress index (CWSI) is a cost-effective, non-destructive, and simple technique for tracking the start of crop water stress. This study investigated the feasibility of CWSI derived from canopy temperature to detect the water status of wheat crops. Artificial intelligence (AI) techniques have become increasingly popular in recent years for determining CWSI. In this study, the performance of two AI techniques, adaptive neuro-fuzzy inference system (ANFIS) and self-organizing maps (SOM), are compared while determining the CWSI of paddy crops. Field experiments were conducted for varying irrigation water applications during two seasons in 2022 and 2023 at the irrigation field laboratory at the Civil Engineering Department, Indian Institute of Technology Roorkee, India. The ANFIS and SOM-simulated CWSI values were compared with the experimentally calculated CWSI (EP-CWSI). Multiple regression analysis was used to determine the upper and lower CWSI baselines. The upper CWSI baseline was found to be a function of crop height and wind speed, while the lower CWSI baseline was a function of crop height, air vapor pressure deficit, and wind speed. The performance of ANFIS and SOM were compared based on mean absolute error (MAE), mean bias error (MBE), root mean squared error (RMSE), index of agreement (d), Nash-Sutcliffe efficiency (NSE), and coefficient of correlation (R²). Both models successfully estimated the CWSI of the paddy crop with higher correlation coefficients and lower statistical errors. However, the ANFIS (R²=0.81, NSE=0.73, d=0.94, RMSE=0.04, MAE= 0.00-1.76 and MBE=-2.13-1.32) outperformed the SOM model (R²=0.77, NSE=0.68, d=0.90, RMSE=0.05, MAE= 0.00-2.13 and MBE=-2.29-1.45). Overall, the results suggest that ANFIS is a reliable tool for accurately determining CWSI in wheat crops compared to SOM.

Keywords: adaptive neuro-fuzzy inference system, canopy temperature, crop water stress index, self-organizing map, wheat

Procedia PDF Downloads 55

Authors: Erwin Chan, Aravind Subramaniyan, Siti Abdullah Fatehah, Steve Lian Kuling

Abstract:

Foam assisted lift technology is proven across the industry to provide efficient deliquification in gas wells. Such deliquification is typically achieved by delivering the foamer chemical downhole via capillary strings. In highly liquid loaded wells where capillary strings are not readily available, foamer can be delivered via batch injection or bull-heading. The latter techniques differ from the former in that cap strings allow for liquid to be unloaded continuously, whereas foamer batches require that periodic batching be conducted for the liquid to be unloaded. Although batch injection allows for liquid to be unloaded in wells with suitable water to gas (WGR) ratio and condensate to gas (CGR) ratio without well intervention for capillary string installation, this technique comes with its own set of challenges - for foamer to de-liquify liquids, the chemical needs to reach perforation locations where gas bubbling is observed. In highly scaled perforation zones in certain wells, foamer delivered in batches is unable to reach the gas bubbling zone, thus achieving poor lift efficiency. This paper aims to discuss the techniques and challenges for unloading liquid via batch injection in scaled perforation wells X and Y, whose WGR is 6bbl/MMscf, whose scale build-up is observed at the bottom of perforation interval, whose water column is 400 feet, and whose ‘bubbling zone’ is less than 100 feet. Variables such as foamer Z dosage, batching technique, and well flow control valve opening times are manipulated during the duration of the trial to achieve maximum liquid unloading and gas rates. During the field trial, the team has found optimal values between the three aforementioned parameters for best unloading results, in which each cycle’s gas and liquid rates are compared with baselines with similar flowing tubing head pressures (FTHP). It is discovered that amongst other factors, a good agitation technique is a primary determinant for efficient liquid unloading. An average increment of 2MMscf/d against an average production of 4MMscf/d at stable FTHP is recorded during the trial.

Keywords: foam, foamer, gas lift, liquid unloading, scale, batch injection

Procedia PDF Downloads 183

Authors: Luís Miguel Cabeça Marques, Alberto Manuel Martinho Vale, José Pedro Miragaia Trancoso Vaz, Ana Sofia Baptista Fernandes, Rui Alexandre de Barros Coito, Tiago Miguel Prates da Costa

Abstract:

The detection and localization of hidden radioactive sources are of significant importance in countering the illicit traffic of Special Nuclear Materials and other radioactive sources and materials. Radiation portal monitors are commonly used at airports, seaports, and international land borders for inspecting cargo and vehicles. However, these equipment can be expensive and are not available at all checkpoints. Consequently, the localization of SNM and other radioactive sources often relies on handheld equipment, which can be time-consuming. The current study presents the advantages of real-time analysis of gamma-ray count rate data from a mobile radiation detection system based on simulated data and field tests. The incorporation of profit functions and decision criteria to optimize the detection system's path significantly enhances the radiation field information and reduces survey time during cargo inspection. For source position estimation, a maximum likelihood estimation algorithm is employed, and confidence intervals are derived using the Fisher information. The study also explores the impact of uncertainties, baselines, and thresholds on the performance of the profit function. The proposed detection system, utilizing a plastic scintillator with silicon photomultiplier sensors, boasts several benefits, including cost-effectiveness, high geometric efficiency, compactness, and lightweight design. This versatility allows for seamless integration into any mobile platform, be it air, land, maritime, or hybrid, and it can also serve as a handheld device. Furthermore, integration of the detection system into drones, particularly multirotors, and its affordability enable the automation of source search and substantial reduction in survey time, particularly when deploying a fleet of drones. While the primary focus is on inspecting maritime container cargo, the methodologies explored in this research can be applied to the inspection of other infrastructures, such as nuclear facilities or vehicles.

Keywords: plastic scintillators, profit functions, path planning, gamma-ray detection, source localization, mobile radiation detection system, security scenario

Procedia PDF Downloads 114

Authors: Siddharth Krishnaswamy

Abstract:

Problem Statement: Currently, when analyzing household food security, the most commonly studied food access indicators are household income and expenditure. Larger studies do take into account other indices such as credit and employment. But these are baselines studies and by definition are conducted infrequently. Food security analysis for access is usually dedicated to analyzing income and expenditure indicators. And both these indicators are notoriously inconsistent. Yet this data can very often end up being the basis on which household food access is calculated; and by extension, be used for decision making. Objectives: This paper argues that along with income and expenditure, credit and debit information should be collected so that an accurate analysis of household food security (and in particular) food access can be determined. The lack of collection and analysis of this information routinely means that there is often a “masking” of the actual situation; a household’s food access and food availability patterns may be adequate mainly as a result of borrowing and may even be due to a long- term dependency (a debt cycle). In other words, such a household is, in reality, worse off than it appears a factor masked by its performance on basic access indicators. Procedures/methodologies/approaches: Existing food security data sets collected in 2005 in Azerbaijan, 2010 across Myanmar and 2014-15 across Uganda were used to support the theory that analyzing income and expenditure of a HHs and analyzing the same in addition to data on credit & borrowing patterns will result in an entirely different scenario of food access of the household. Furthermore, the data analyzed depicts food consumption patterns across groups of households and then relates this to the extent of dependency on credit, i.e. households borrowing money in order to meet food needs. Finally, response options that were based on analyzing only income and expenditure; and response options based on income, expenditure, credit, and borrowing – from the same geographical area of operation are studied and discussed. Results: The purpose of this work was to see if existing methods of household food security analysis could be improved. It is hoped that food security analysts will collect household level information on credit and debit and analyze them against income, expenditure and consumption patterns. This will help determine if a household’s food access and availability are dependent on unsustainable strategies such as borrowing money for food or undertaking sustained debts. Conclusions: The results clearly show the amount of relevant information that is missing in Food Access analysis if debit and borrowing of the household is not analyzed along with the typical Food Access indicators that are usually analyzed. And the serious repercussions this has on Programmatic response and interventions.

Keywords: analysis, food security indicators, response, resilience analysis

Procedia PDF Downloads 331

Authors: Simon Espigolé, Igor Shalyminov, Helen Hastie

Abstract:

Deep machine learning techniques including deep neural networks (DNN) have been used to model language and dialogue for conversational agents to perform tasks, such as giving technical support and also for general chit-chat. They have been shown to be capable of generating long, diverse and coherent sentences in end-to-end dialogue systems and natural language generation. However, these systems tend to imitate the training data and will only generate the concepts and language within the scope of what they have been trained on. This work explores how deep neural networks can be used in a task that would normally require human creativity, whereby the human would read the movie description and/or watch the movie and come up with a compelling, interesting movie title. This task differs from simple summarization in that the movie title may not necessarily be derivable from the content or semantics of the movie description. Here, we train a type of DNN called a sequence-to-sequence model (seq2seq) that takes as input a short textual movie description and some information on e.g. genre of the movie. It then learns to output a movie title. The idea is that the DNN will learn certain techniques and approaches that the human movie titler may deploy that may not be immediately obvious to the human-eye. To give an example of a generated movie title, for the movie synopsis: ‘A hitman concludes his legacy with one more job, only to discover he may be the one getting hit.’; the original, true title is ‘The Driver’ and the one generated by the model is ‘The Masquerade’. A human evaluation was conducted where the DNN output was compared to the true human-generated title, as well as a number of baselines, on three 5-point Likert scales: ‘creativity’, ‘naturalness’ and ‘suitability’. Subjects were also asked which of the two systems they preferred. The scores of the DNN model were comparable to the scores of the human-generated movie title, with means m=3.11, m=3.12, respectively. There is room for improvement in these models as they were rated significantly less ‘natural’ and ‘suitable’ when compared to the human title. In addition, the human-generated title was preferred overall 58% of the time when pitted against the DNN model. These results, however, are encouraging given the comparison with a highly-considered, well-crafted human-generated movie title. Movie titles go through a rigorous process of assessment by experts and focus groups, who have watched the movie. This process is in place due to the large amount of money at stake and the importance of creating an effective title that captures the audiences’ attention. Our work shows progress towards automating this process, which in turn may lead to a better understanding of creativity itself.

Keywords: creativity, deep machine learning, natural language generation, movies

Procedia PDF Downloads 326

Authors: B. Orser, A. Riding, Y. Li

Abstract:

To inform government about how gender gaps in SME ( small and medium-sized enterprise) contracting might be redressed, the research question was: What are the key obstacles to, and response strategies for, increasing the engagement of women business owners among SME suppliers to the government of Canada? Thirty-five interviews with senior policymakers, supplier diversity organization executives, and expert witnesses to the Canadian House of Commons, Standing Committee on Government Operations and Estimates. Qualitative data were conducted and analysed using N’Vivo 11 software. High order response categories included: (a) SME risk mitigation strategies, (b) SME procurement program design, and (c) performance measures. Primary obstacles cited were government red tape and long and complicated requests for proposals (RFPs). The majority of 'common' complaints occur when SMEs have questions about the federal procurement process. Witness responses included use of outcome-based rather than prescriptive procurement practices, more agile procurement, simplified RFPs, making payment within 30 days a procurement priority. Risk mitigation strategies included provision of procurement officers to assess risks and opportunities for businesses and development of more agile procurement procedures and processes. Recommendations to enhance program design included: improved definitional consistency of qualifiers and selection criteria, better co-ordination across agencies; clarification about how SME suppliers benefit from federal contracting; goal setting; specification of categories that are most suitable for women-owned businesses; and, increasing primary contractor awareness about the importance of subcontract relationships. Recommendations also included third-party certification of eligible firms and the need to enhance SMEs’ financial literacy to reduce financial errors. Finally, there remains the need for clear and consistent pre-program statistics to establish baselines (by sector, issuing department) performance measures, targets based on percentage of contracts granted, value of contract, percentage of target employee (women, indigenous), and community benefits including hiring local employees. The study advances strategies to enhance federal procurement programs to facilitate socio-economic policy objectives.

Keywords: procurement, small business, policy, women

Procedia PDF Downloads 113

Authors: Julia Ive, Lucia Specia

Abstract:

Understanding and detecting toxicity is an important problem to support safer human interactions online. Our work focuses on the important problem of contextual toxicity detection, where automated classifiers are tasked with determining whether a short textual segment (usually a sentence) is toxic within its conversational context. We use “toxicity” as an umbrella term to denote a number of variants commonly named in the literature, including hate, abuse, offence, among others. Detecting toxicity in context is a non-trivial problem and has been addressed by very few previous studies. These previous studies have analysed the influence of conversational context in human perception of toxicity in controlled experiments and concluded that humans rarely change their judgements in the presence of context. They have also evaluated contextual detection models based on state-of-the-art Deep Learning and Natural Language Processing (NLP) techniques. Counterintuitively, they reached the general conclusion that computational models tend to suffer performance degradation in the presence of context. We challenge these empirical observations by devising better contextual predictive models that also rely on NLP data augmentation techniques to create larger and better data. In our study, we start by further analysing the human perception of toxicity in conversational data (i.e., tweets), in the absence versus presence of context, in this case, previous tweets in the same conversational thread. We observed that the conclusions of previous work on human perception are mainly due to data issues: The contextual data available does not provide sufficient evidence that context is indeed important (even for humans). The data problem is common in current toxicity datasets: cases labelled as toxic are either obviously toxic (i.e., overt toxicity with swear, racist, etc. words), and thus context does is not needed for a decision, or are ambiguous, vague or unclear even in the presence of context; in addition, the data contains labeling inconsistencies. To address this problem, we propose to automatically generate contextual samples where toxicity is not obvious (i.e., covert cases) without context or where different contexts can lead to different toxicity judgements for the same tweet. We generate toxic and non-toxic utterances conditioned on the context or on target tweets using a range of techniques for controlled text generation(e.g., Generative Adversarial Networks and steering techniques). On the contextual detection models, we posit that their poor performance is due to limitations on both of the data they are trained on (same problems stated above) and the architectures they use, which are not able to leverage context in effective ways. To improve on that, we propose text classification architectures that take the hierarchy of conversational utterances into account. In experiments benchmarking ours against previous models on existing and automatically generated data, we show that both data and architectural choices are very important. Our model achieves substantial performance improvements as compared to the baselines that are non-contextual or contextual but agnostic of the conversation structure.

Keywords: contextual toxicity detection, data augmentation, hierarchical text classification models, natural language processing

Procedia PDF Downloads 169

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: Cynthia Sau Chun Yip, Richard Fielding

Abstract:

High meat and especially red meat intakes are significantly and positively associated with a multiple burden of diseases and also high greenhouse gas (GHG) emissions. This study investigated population meat intake patterns in Hong Kong. It quantified the burden of disease and GHG emission outcomes by modeling to adjust Hong Kong population meat intakes to recommended healthy levels. It compared age- and sex-specific population meat, fruit and vegetable intakes obtained from a population survey among adults aged 20 years and over in Hong Kong in 2005-2007, against intake recommendations suggested in the Modelling System to Inform the Revision of the Australian Guide to Healthy Eating (AGHE-2011-MS) technical document. This study found that meat and meat alternatives, especially red meat intakes among Hong Kong males aged 20+ years and over are significantly higher than recommended. Red meat intakes among females aged 50-69 years and other meat and alternatives intakes among aged 20-59 years are also higher than recommended. Taking the 2005-07 age- and sex-specific population meat intake as baselines, three counterfactual scenarios of adjusting Hong Kong adult population meat intakes to AGHE-2011-MS and Pre-2011 AGHE recommendations by the year 2030 were established. Consequent energy intake gaps were substituted with additional legume, fruit and vegetable intakes. To quantify the consequent GHG emission outcomes associated with Hong Kong meat intakes, Cradle-to-ready-to-eat lifecycle assessment emission outcome modelling was used. Comparative risk assessment of burden of disease model was used to quantify the health outcomes. This study found adjusting meat intakes to recommended levels could reduce Hong Kong GHG emission by 17%-44% when compared against baseline meat intake emissions, and prevent 2,519 to 7,012 premature deaths in males and 53 to 1,342 in females, as well as multiple burden of diseases when compared to the baseline meat intake scenario. Comparing lump sum meat intake reduction and outcome measures across the entire population, and using emission factors, and relative risks from individual studies in previous co-benefit studies, this study used age- and sex-specific input and output measures, emission factors and relative risks obtained from high quality meta-analysis and meta-review respectively, and has taken government dietary recommendations into account. Hence evaluations in this study are of better quality and more reflective of real life practices. Further to previous co-benefit studies, this study pinpointed age- and sex-specific population and meat-type-specific intervention points and leverages. When compared with similar studies in Australia, this study also showed that intervention points and leverages among populations in different geographic and cultural background could be different, and that globalization also globalizes meat consumption emission effects. More regional and cultural specific evaluations are recommended to promote more sustainable meat consumption and enhance global food security.

Keywords: burden of diseases, greenhouse gas emissions, Hong Kong diet, sustainable meat consumption

Procedia PDF Downloads 311

Authors: Chia-Pei Chen, Yu-Ju Chen, Yu-Juei Hsu

Abstract:

The prevalence and incidence of end-stage renal disease in Taiwan ranks the highest in the world. According to the statistical survey of the Ministry of Health and Welfare in 2019, kidney disease is the ninth leading cause of death in Taiwan. It leads to autonomic dysfunction, inflammatory response and symptom distress, and further increases the damage to the structure and function of the kidneys, leading to increased demand for renal replacement therapy and risks of cardiovascular disease, which also has medical costs for the society. If we can intervene in a feasible manual to effectively regulate the autonomic nerve function of CKD patients, reduce the inflammatory response and symptom distress. To prolong the progression of the disease, it will be the main goal of caring for CKD patients. This study aims to test the effect of heart rate variability biofeedback (HRVBF) on improving autonomic nerve function (Heart Rate Variability, HRV), inflammatory response (Interleukin-6 [IL-6], C reaction protein [CRP] ), symptom distress (Piper fatigue scale, Pittsburgh Sleep Quality Index [PSQI], and Beck Depression Inventory-II [BDI-II] ) in patients with chronic kidney disease. This study was experimental research, with a convenience sampling. Participants were recruited from the nephrology clinic at a medical center in northern Taiwan. With signed informed consent, participants were randomly assigned to the HRVBF or control group by using the Excel BINOMDIST function. The HRVBF group received four weekly hospital-based HRVBF training, and 8 weeks of home-based self-practice was done with StressEraser. The control group received usual care. We followed all participants for 3 months, in which we repeatedly measured their autonomic nerve function (HRV), inflammatory response (IL-6, CRP), and symptom distress (Piper fatigue scale, PSQI, and BDI-II) on their first day of study participation (baselines), 1 month, and 3 months after the intervention to test the effects of HRVBF. The results were analyzed by SPSS version 23.0 statistical software. The data of demographics, HRV, IL-6, CRP, Piper fatigue scale, PSQI, and BDI-II were analyzed by descriptive statistics. To test for differences between and within groups in all outcome variables, it was used by paired sample t-test, independent sample t-test, Wilcoxon Signed-Rank test and Mann-Whitney U test. Results: Thirty-four patients with chronic kidney disease were enrolled, but three of them were lost to follow-up. The remaining 31 patients completed the study, including 15 in the HRVBF group and 16 in the control group. The characteristics of the two groups were not significantly different. The four-week hospital-based HRVBF training combined with eight-week home-based self-practice can effectively enhance the parasympathetic nerve performance for patients with chronic kidney disease, which may against the disease-related parasympathetic nerve inhibition. In the inflammatory response, IL-6 and CRP in the HRVBF group could not achieve significant improvement when compared with the control group. Self-reported fatigue and depression significantly decreased in the HRVBF group, but they still failed to achieve a significant difference between the two groups. HRVBF has no significant effect on improving the sleep quality for CKD patients.

Keywords: heart rate variability biofeedback, autonomic nerve function, inflammatory response, symptom distress, chronic kidney disease

Procedia PDF Downloads 180

Authors: Swaleh Aboud, Mishal Gudka, David Obura

Abstract:

Background: Coral reefs are becoming increasingly vulnerable due to several threats ranging from climate change to overfishing. This has resulted in increased management and conservation efforts to protect reefs from degradation and facilitate recovery. Recruitmentof new individuals are isimportant in the recovery process and critical for the persistence of coral reef ecosystems. Local coral community structure can be influenced by successful recruit settlement, survival, and growth Understanding coral recruitment patterns can help quantify reef resilience and connectivity, establish baselines and track changes and evaluate the effectiveness of reef restoration and conservation efforts. This study will examine the abundance and spatial pattern of coral recruits and how this relates to adult community structure, including the distribution of thermal resistance and sensitive genera and their distribution in different management regimes. Methods: Coral recruit and demography surveys were conducted from 2020 to 2022, covering 35 sites in 19coral reef locations along the Kenyan coast. These included marine parks, reserves, community conservation areas (CMAs), and open access areas from the north (Marereni) to the south (Kisite) coast of Kenya and across different reef habitats. The data was collected through the underwater visual census (UVC) technique. We counted adult corals (>10 cm diameter)of23 selected genera using belt transects (25 by 1 m) and sampling of 1 m2 quadrat (at an interval of 5m) for all coloniesless than 10 cm diameter. The benthic cover was collected using photo quadrats. The surveys were only done during the northeast monsoon season. The data wereanalyzed using the R program to see the distribution patterns and the Kruskal Wallis test to see whether there was a significant difference. Spearman correlation was also applied to assess the relationship between the distribution of coral genera in recruits and adults. Results: A total of 44 different coral genera were recorded for recruits, ranging from 3at Marereni to 30at Watamu Marine Reserve. Recruit densities ranged from 1.2±1.5recruit m-2 (mean±SD) at Likoni to 10.3± 8.4 recruit m-2 at Kisite Marine Park. The overall densityof recruitssignificantly differed between reef locations, with Kisite Marine Park and Reserve and Likonihaving significantly large differences from all the other locations, while Vuma, Watamu, Malindi, and Kilifi had significantly lower differences from all the other locations. The recruit generadensity along the Kenya coastwas divided into two clusters, one of which only included sites inKisite Marine Park. Adult colonies were dominated by Porites massive, Acropora, Platygyra, and Favites, whereas recruits were dominated by Porites branching, Porites massive, Galaxea, and Acropora. However, correlation analysis revealed a statistically significant positive correlation (r=0.81, p<0.05) between recruit and adult coral densities across the 23 coral genera. Marereni, which had the lowest densityof recruits, has only thermallyresistant coral genera, while Kisite Marine Park, with the highest recruit densities, has over 90% thermal sensitive coral genera. A weak positive correlation was found between recruit density and coralline algae, dead standing corals, and turf algae, whereas a weak negative correlation was found between recruit density and bare substrate and macroalgae. Between management regimes, marine reserves were found to have more recruits than no-take zones (marine parks and CMAs) and open access areas, although the difference was not significant. Conclusion: There was a statistically significant difference in the density of recruits between different reef locations along the Kenyan coast. Although the dominating genera of adults and recruits were different, there was a strong positive correlation between their coral communities, which could indicate self-recruitment processes or consistent distance seedings (of the same recruit genera). Sites such as Kisite Marine Park, with high recruit densities but dominated by thermally sensitive genera, will, on the other hand, be adversely affected by future thermal stress. This could imply that reducing the threats to coral reefs such as overfishingcould allow for their natural regeneration and recovery.

Keywords: coral recruits, coral adult size-class, cora demography, resilience

Procedia PDF Downloads 124