Search

This paper focuses on a comparison study between the traditional deep learning techniques like LSTMs and GRUs with the latest state-of-the-art transformers for the task of classifying a comment based on its toxicity. Specifically, four different deep learning models are built, implemented and trained on a standard dataset for comparing the performance of the models. The models that are explored and compared are Bi-directional LSTMs, GRUs, Bidirectional LSTMs with CNNs and Transformers (with pre-trained RoBERTa weights). [ABSTRACT FROM AUTHOR]

There are many problems related to image processing and analysis that could be solved using convolutional neural networks (CNN). It's easy to implement CNN using one of thousand high-level programming languages. Such CNN will not be fast and energy efficient enough to be used in real-time systems. The good way to solve this problem is to use special hardware accelerators (neuroprocessors). The paper shows that it is possible to reduce the calculation time of the network using neuroprocessors. The implementation of such hardware is quite challenging process that requires specialized knowledge. That's why we need a tool for automated hardware synthesis. The basis of this tool are various optimizations that will allow us to transfer the network to a hardware platform. The optimization mechanisms in existing tools are either very poor or nonexistent. We propose several optimizations on different stages of developing process of target hardware. In the paper the authors describe a CNN model for handwritten digits recognition and show how to reduce the number of neural network parameters without significant accuracy losses. The authors managed to reduce the number of parameters from 644 120 to 31 530 with accuracy loss just about 0.43%, making the CNN suitable for synthesis on dedicated hardware platform. The authors also examined the dependence of the target platform resources on the method of computing the neural network output (sequentially / pipelined / parallel). It was showed that it is possible to decrease computation time in 7 times using fully parallel computations, bit it required in 4-6 times more resources than using sequential calculations. Using the above, as well as many other optimizations, it allows to create a tool for automated synthesis of high-quality hardware accelerators for CNN. The paper also presents the concept of such tool. [ABSTRACT FROM AUTHOR]

Parkinson’s disease is a tragedy for retirees because it causes the nervous framework to deteriorate and, as a consequence, the participant's maneuverability. The application's precise involvement in Parkinson's disease significantly increases the chances of preventing it from worsening. The framework can circumvent this by gathering exact data from the individual or by supplying certain sample handwritten patterns, such as spirals or waves. This review article looked at a collection of appropriate publications as a solution. The research demonstrates that the computer vision architecture is the final answer for revealing the Parkinson disease probability by studying the sufferer's hand-drawn image. As a response, relevant papers have been studied in attempt to create an effective recognition system that employs the Region of Interest model to recognize the fuzzy pattern of patients' symptoms. The spiral shapes and accompanying discrepancies are caught using the CNN and Decision Making procedures in the Classification models. Future editions of this study will go through the approach in greater detail. [ABSTRACT FROM AUTHOR]

By taking advantage of their nonlinear dynamics Cellular Nonlinear Networks (CNN) are considered to be powerful tools for many image processing applications. In this paper we will try to investigate the feasibility of image representation by using CNN dynamics. [ABSTRACT FROM PUBLISHER]

Road accidents due to distracted driving have been on a rise in recent years. As per the Road Accident Report 2019, 11 people were killed each hour in India due to road accidents (Transport Research Wing, 2020). This makes it crucial to take measures to stop the number of road fatalities. It found that the major cause of these accidents was driver error. This paper proposed a solution to detect the distraction of drivers into different predefined classes. The use of different pre-trained Convolutional Neural Network (CNN) models viz. AlexNet, VGG16, and ResNet50, for the classification of distracted drivers according to the State Farm's Distracted Driver Detection challenge on Kaggle, are depicted in this paper, As well as MyModel is trained on Dataset consisting of different local driver's 2D dashboard camera images along with the State Farm's Dataset. After comparing the results of predefined models with MyModel, the best result has been found as a categorical cross-entropy loss of 0.2344 on the validation set and an accuracy score of 93.28%. [ABSTRACT FROM AUTHOR]

The multi-model image registration is widely used in medical clinical diagnosis. However, the feature descriptor pairs are hard to detect between different modalities, which is a common dilemma faced in multi-modal image registration. In this paper, we detect feature descriptor pairs in the structure representation by convolutional neural network (CNN). On the one hand, structure representation can transform different modalities into a third-type modality, and it shows the potential information which do not appear at multi-modal medical image commonly. On the other hand, the matched feature points pairs which are computed by CNN can move back from structure representation to original image. Finally, the effective feature points are matched by 2NN algorithm. Extensive experiments on group-wise registration prove that this algorithm overcome the dilemma in extracting feature descriptor pairs among the multi-modal medical images. [ABSTRACT FROM AUTHOR]

With the development of society, the proportion of sitting time in people's daily life and working environment is increasing. However, Most of the sitting posture is improper sitting posture in most of the time. Improper sitting posture for a long time is one of the main causes of a series of skeletal muscle diseases. In this paper, a CNN sitting posture recognition model based on human pressure data is proposed. The model is constructed by collecting a large number of pressure data of human-chair contact surface and using these training to get sitting posture recognition algorithm. Experiments show that the algorithm can accurately identify eight kinds of human sitting posture. Its accuracy rate is as high as 95.6%, and the recall rate is 95.5% at same time. The sitting position recognition system constructed by this algorithm can monitor and distinguish the bad sitting posture of human body in real time, and it has the many advantages such as high accuracy, high robustness, high availability and high security [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNN) achieve great success in computer vision, it can learn hierarchical representation from raw pixels and has outstanding performance in various image recognition tasks [1]. However, CNN is easy to be fraudulent in terms of it is possible to produce images totally unrecognizable to human eyes that CNNs believe with near certainty are familiar objects. [2]. In this paper, an associative memory model based on multiple features is proposed. Within this model, feature extraction and classification are carried out by CNN, T-SNE and exponential bidirectional associative memory neural network (EBAM). The geometric features extracted from CNN and the digital features extracted from T-SNE are associated by EBAM. Thus we ensure the recognition of robustness by a comprehensive assessment of the two features. In our model, we can get only 8% error rate with fraudulent data. In systems that require a high safety factor or some key areas, strong robustness is extremely important, if we can ensure the image recognition robustness, network security will be greatly improved and the social production efficiency will be extremely enhanced. [ABSTRACT FROM AUTHOR]

Smartwatches have the potential to provide glanceable, always-available sound feedback to people who are deaf or hard of hearing. In this paper, we present a performance evaluation of four low-resource deep learning sound classification models: MobileNet, Inception, ResNet-lite, and VGG-lite across four device architectures: watch-only, watch+phone, watch+phone+cloud, and watch+cloud. While direct comparison with prior work is challenging, our results show that the best model, VGG-lite, performed similar to the state of the art for non-portable devices with an average accuracy of 81.2% (SD=5.8%) across 20 sound classes and 97.6% (SD=1.7%) across the three highest-priority sounds. For device architectures, we found that the watch+phone architecture provided the best balance between CPU, memory, network usage, and classification latency. Based on these experimental results, we built and conducted a qualitative lab evaluation of a smartwatch-based sound awareness app, called SoundWatch (Figure 1), with eight DHH participants. Qualitative findings show support for our sound awareness app but also uncover issues with misclassifications, latency, and privacy concerns. We close by offering design considerations for future wearable sound awareness technology. [ABSTRACT FROM AUTHOR]

Optimizers are approaches or algorithms dependent to enhance the characteristics of the Neural Network (NN) like weights and learning rate in order to decrease the loss rate, On the other hand, Siamese Neural Network (SNN) are two identical sub-networks, they work in parallel and they are sharing parameters and weight, SNN uses for indicate similarity. In this research, we study the effect of optimizers Siamese Neural Network, using Digits handwritten (MINST) dataset, the effects is studied for Adam, Nadam, Adadelta and SGD optimizers with respect to process time and accuracy, the accuracy is 97%, 97%, 79% and 92%. [ABSTRACT FROM AUTHOR]

In this paper, we propose an automated framework that takes as input a TensorFlow inference graph and generates high-performance accelerators on FPGA by assembling CNN pre-implemented components as a puzzle, based on the graph topology. Using pre-implemented components allows us the only use the minimum of resources necessary, predict the performance and a gain in productivity We adopt a unified representation based on systolic array to perform the computational-hungry operations of the model and provide novel analysis of design trade-offs for FPGA CNN accelerators. Experimental results show the great performance, low latency and flexibility provided by the proposed framework. [ABSTRACT FROM AUTHOR]

This study investigates the decade long effort to construct and validate a communications concept of international relations, which asserts that global television networks, such as CNN and BBC World, have become the decisive actor in determining policies and outcomes of significant events. The article traces the origins and development of this concept, known also as the CNN effect, and critically analyzes major academic and professional publications that have sought to explain it. These publications include theoretical and comparative works, specific case studies, and methodologies. The study reveals an ongoing debate among politicians, government officials, journalists, and scholars, on the validity of this theory. This article concludes that studies have yet to present sufficient evidence validating the CNN effect, that many works have exaggerated this effect, and that the focus on this concept has deflected attention from other ways global television affects mass communication, journalism, and international relations. The paper also proposes an agenda for research on the various effects of global television networks. [ABSTRACT FROM AUTHOR]

This paper applies rational choice theory to help explain why two transnational television news networks, CNN and Al-Jazeera, made sharply contrasting decisions about airing a videotaped interview with Osama bin Laden in the months following the September 11 attacks. Both organizations sought to justify their decisions in terms of professional norms, particularly in expressing their autonomy of news judgment. However, as geopolitical actors, the networks faced far different structural considerations. Al-Jazeera, which still depended on the sponsorship of the emir of Qatar, had to weigh the consequences of its decision on its reputation as an emerging independent network. The networks' varied strategies reveal the gamesmanship inherent in global mass communication and suggest as well that audiences are inheriting the potential for more complex understandings. [ABSTRACT FROM AUTHOR]

The life cycles of current electronic devices are getting shorter due to the multitude of innovations, resulting in a significant amount of waste of electrical and electronic equipment (WEEE). This electronic scrap contains many valuable materials, in particular metals such as gold, silver, platinum or even tantalum. While some of them are already being recycled, there is still much room left for optimizations. One possibility in order to optimize the current recycling process of rare and valuable metals is the selective disassembly of certain components. This leads to a higher concentration of the desired substances, enabling an economical recycling. In order to disassemble the components selectively, they have to be detected in the first place. A pilot plant with different sensors has been built for this purpose. The plant delivers preprocessed data, which is fed into feature extraction and classification routines. Based on previous research this paper focuses on a comparison between two different feature-based methods and a state-of-the-art approach in the form of a convolutional neural network for detection and identification of electronic components containing valuable substances. The feature-based methods consist of a Random Forest (RF) classifier in combination with either a Principal Component Analysis (PCA) or a predefined feature descriptor. [ABSTRACT FROM AUTHOR]

In January 2016, Iran detained ten U.S. sailors who had drifted into Iranian territory, releasing them fifteen hours later after intense diplomatic activity. This paper explores the discursive choices evident over three days of broadcast and cable coverage to illuminate takenfor-granted ideological stances common to several of the reports, with particular emphasis on two clips from CNN and Fox News that bear significant formal distinctions concealing close substantive similarities. The analysis calls into question dominant views on the ideological positioning of cable networks and the discursive emphases among varying journalistic formats, with implications for U.S. domestic politics and global relations. [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNNs) have become the state-of-the-art technique in many classification tasks in IoT system. However, the low-power and area-constraint edge devices are unable to afford the expensive cost of CNNs. Resistive random access memory (RRAM) is attractive for establishing the CNN accelerator at the edge end due to the features of scalability, low-power and in-situ dot- product. However, mapping a random network architecture onto a general-purpose RRAM accelerator suffers a severe issue of resource underutilization. The neural network quantization offers an opportunity to rescue the degraded resource utilization. Selecting the bit-width for the vast parameters is impractically completed by human labor. This paper proposes an AutoML-based array-aware quantization and mapping framework that generates the fine-grained mixed-precision neural networks to optimize resource utilization in RRAM. In this framework, we design a two-stage learning and array- aware grouping strategy to quickly explore the huge searching space. The experimental results show that the proposed framework achieves 18.2%~36.1% improvement in resource utilization and 0.9%~3.3% increase in model accuracy over prior coarse-grained quantization methods. [ABSTRACT FROM AUTHOR]

Repeat pattern is a common structure in textile or graphic design, while scalable and reusable designs are required for production. We propose a new approach, based on Convolutional Neural Network (CNN), for automatic extraction of repeat patterns in this paper. CNNs are good at detecting repeat patterns through extracting multilevel features in the convolutional layers. In our method, we select filters quickly in single images by leveraging the combination of the learned filters from AlexNet’s convolutional layers and boundary detection results from VGG16. Moreover, we use template matching to optimize final outputs in order to improve precision. We composed a dataset with 30 images, covering stripe, check and dot distributed patterns, each image with manual ground truth labels. The experimental results show that our method outperforms the state-of-the-art method in both precision and running speeds. [ABSTRACT FROM AUTHOR]

According to the data of the Turkish Public Health Institution Cancer Department, cancer is seen in 161 thousand people annually in our country. Malignant tumor is called cancer. Brain tumor disease is a common disease in our country. Brain tumors are growing fast. The time to be noticed and, accordingly, to start treatment is important. If there is a tumor in the brain, its type (benign, malignant) should be determined. Early diagnosis is vital. This process, which is determined by experts, can take a long time and there is a possibility of making mistakes. For this reason, by including artificial intelligence methods in the detection process of brain tumors, the time can be shortened and the probability of error can be reduced. For this purpose, a study was conducted using Convolutional Neural Networks, one of the deep learning algorithms. In the study, 6083 MR images were used. The data set was primarily classified according to whether there was a brain tumor and was divided into 70% training and 30% testing. These classified images are binarized in order to train the model comfortably. CNN deep learning model was applied to the pre-processed data set and the accuracy rate was found to be 99%. [ABSTRACT FROM AUTHOR]

In this paper a DTCNN model for dynamic object segmentation in videos is presented. The proposed method involves three main stages; dynamic background registration, dynamic objects detection and object segmentation improvement. Two DTCNNs are used, one to achieved object detection and other for morphologic operations in order to improve object segmentation. Visual and quantitative results are compared with findings of a Self-organizing map SOM-like dynamic object detection approach. Considering the experiments reported, it can be said that the proposed method shows acceptable results with some improvements over the SOM because the DTCNN method does not need human intervention for parameter adjustment. [ABSTRACT FROM PUBLISHER]

In today's media environment, people have innumerable choices for obtaining political news and information. This study investigates what underlies people's cable news choices by analyzing their perceptions of FOX and CNN. Using data gathered from the 2006 Cooperative Congressional Election Study, this paper confirms that many people have chosen sides, finding one of these cable news networks to be far more trustworthy than the other. These individuals have demographic and political characteristics that distinguish them from others who trust both CNN and FOX and from others who distrust both CNN and FOX. This study also finds that trust mediates the relationship between people's political predispositions and their cable news exposure habits. Both trusting one's preferred outlet and distrusting an alternative outlet motivate people's exposure patterns. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]

This study explored different hyperparameters and their outcomes on multiple DNN architectures by incorporating transfer learning technique on the plant seedling dataset. Optimizers can have different strengths and weaknesses, and their performance may depend on the specific characteristics of the model and the dataset being used. We have observed that the choice of the optimizer is just one of many hyperparameters that can affect the performance of deep neural network (DNN) models. Other hyperparameters, such as the droprate, number of epochs, and batch size, can also have a significant impact. MobileNetV2 demonstrated superior performance while maintaining a smaller model size, making it a highly valuable option for edge devices where size is a crucial factor. [ABSTRACT FROM AUTHOR]

Targeting reliable Covid-19 detection, we report the results of a comparative analysis focused on using three distinct types of spatial representations of cough recordings applied as inputs to Convolutional Neural Networks (CNN’s). The actual algorithms used to transform the original time series into images are based on the classical Mel Frequency Cepstral Coefficients (MFCC) spectrogram, a modified version of the Continuous Wavelet Transform (the S-Transform), and the Gramian Angular Fields (GAF). Extensive experiments have been conducted on the COUGHVID and COVID-19 Sounds cough datasets, using various CNN’s architectures. The MFCC based approach yielded best results, comparing favorably to existing reported performances. [ABSTRACT FROM AUTHOR]

Breast cancer, the most common of all the cancers, is treatable when detected early. Histopathology (HP) biopsy images generated from breast tissue samples are commonly used as early screening tools for detecting malignancy. This study has developed and compared the efficacy of two convolutional neural network (CNN) models that are based on MobileNet architecture for automatic detection of Invasive Ductal Carcinoma (IDC), the most common form of breast cancer, in histopathology (HP) images of breast tissues. The best of the two models has shown encouraging classification performance in terms of accuracy, precision and recall. The results attest to the viability of using lighter CNN models such as MobileNet for decision support when screening for breast cancer. [ABSTRACT FROM AUTHOR]

This study examines the structure of international aid flows in the global system and the relationship between news coverage and international aid allocations from a communication perspective. Generally, traditional perspectives on international aid have focused on the nation-state system, limiting explanation of international aid flows in the global system. World systems theory is proposed to explain the phenomenon. The results of a network analysis found: 1) international aid flows follow a core-periphery structure; 2) the relationships between donors and recipients reflect the structural characteristics of international interaction described by Galtung (1971). This study also suggests the bystander intervention model to explain the mechanism of international aid. The model emphasizes the importance of communication among members in a social system. It represents the processes of decision-making on a nationÂ’s policy to help other countries and provides the mechanism of international communication that encourages aid. The global news media plays an important role as an agenda-setting agent. The results of multiple regression analysis reveal that global news media coverage is a significant determinant of international aid. ..PAT.-Conference Proceeding [ABSTRACT FROM AUTHOR]

The article provides information on the findings of the survey conducted on Al-Jazeera television viewers through a survey posted on the network's Arab language Web site to examine how credible Al-Jazeera views judge the network. More specifically, the article examines the degree to which Al-Jazeera watchers rate the station on depth, accuracy, fairness, believability, trustworthiness and expertise.

The article examines the transcripts of several episodes of "Reliable Sources" by using the theory of hegemony of Antonio Gramsci to illustrate how the program avoids addressing the problems inherent in the U.S. news media. The form of media criticism depicted in the television program serves the interests of those who control and operate the media as well as professional journalists.

This framing analysis of CNN?s live coverage of September 11 provides some insight into how broadcast media framed the events both verbally and visually. A verbal (audio) and visual framing analysis of the first 12 hours of ?breaking? live news broadcast by CNN was conducted. This analysis qualitatively deconstructs and analyzes CNN?s verbal and visual ?breaking news? coverage and suggests that CNN created a powerful frame with its coverage that argued to viewers that the events of September 11 comprised an act of war so horrific that immediate military retaliation was not only justified but necessary. [ABSTRACT FROM AUTHOR]

This study investigates the effects of global television news on the formulation and implementation of U.S. foreign policy. While it found no evidence to support the "CNN effect," a hypothesis that claims global television now determines policy, it does present evidence and analysis of other significant effects on various phases and dimensions of policymaking. This study also suggests that global news coverage has created political, professional, and ethical dilemmas for the three major groups of participants in policymaking: politicians, officials, and journalists. Furthermore, it asserts that these groups have not yet sufficiently adjusted to the new television saturated environment of policymaking. [ABSTRACT FROM AUTHOR]

Copyright of Congreso Internacional de Investigacion Academia Journals is the property of PDHTech, LLC and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The unstructured sparsity after pruning poses a challenge to the efficient implementation of deep learning models in existing regular architectures like systolic arrays. The coarse-grained structured pruning, on the other hand, tends to have higher accuracy loss than unstructured pruning when the pruned models are of the same size. In this work, we propose a compression method based on the unstructured pruning and a novel weight permutation scheme. Through permutation, the sparse weight matrix is further compressed to a small and dense format to make full use of the hardware resources. Compared to the state-of-the-art works, the matrix compression rate is effectively improved from 5.88x to 10.28x. As a result, the throughput and energy efficiency are improved by 2.12 and 1.57 times, respectively. [ABSTRACT FROM AUTHOR]

Prior research on energy-efficient Convolutional Neural Network (CNN) inference accelerators mostly focus on exploiting the model sparsity, i.e., zero patterns in weight and activations, to reduce the on-chip storage and computation overhead. In this work, we found in addition to zero patterns, a larger group of repetitive patterns and values exists in the working-set of CNN inference task, which is defined as computation redundancy and induces unnecessary performance and storage overhead in CNN accelerators. Based on this observation, we proposed a redundancy-free architecture that detects and eliminates the repetitive computation and storage patterns in CNN for more efficient network inference. The architecture consists of two parts: the off-line parameter analyzer that extracts the repetitive patterns in the 3D tensor of parameters, and the dataflow accelerator. The proposed accelerator at first preprocesses the weight patterns and the dynamically generated activations, and then cache these intermediate results in special P²-cache banks for further usage in convolution or full-connection stage. It is evaluated in experiments that the proposed Cavoluche architecture removes up to 89% of the repetitive operations from the layer inference process and reduce 77% of on-chip storage space to store both redundancy-free weight and activations. It is seen in experiments that the implementation of Cavoluche outperforms the state-of-the-art mobile GPGPU in both performance and energy-efficiency. When compared to the latest sparsity base accelerators, Cavoluche also achieves better operation elimination effects. [ABSTRACT FROM AUTHOR]

Deep convolution neural networks (CNN) have demonstrated remarkable performance on visual recognition tasks. This performance is feasible due to availability of massive datasets used for training of the networks, and requires long training time even on fast GPUs. One of the challenges related to application of CNN for image recognition is training a new class/category into the model. This usually requires that training data for the already learned classes is available to be used for re-training the model with the training examples for the new class. In this work we propose a modular architecture of image recognition based on n CNNs trained to recognize n individual image categories rather than train one CNN with n-way Soft Max as the last layer. We empirically show that this approach offers comparable performance with state-of-the-art CNNs, while allowing for much easier expansion in terms of classes learned by the model as compared with standard models based on a single CNN with an n-way Soft Max. [ABSTRACT FROM AUTHOR]

The study compared two versions (Arabic and English) of two multi-language Web sites, CNN and Al-Jazeera, seeking an understanding of how news organizations adapt to different cultural settings by changing the parameters that shape their agendas.The study found that Web sites targeting similar audiences correlated in terms of the first-level agenda - topics that are of interest to their common target audience. However, each edition - Arabic or English - of each Web site - CNN or Al-Jazeera -reflected the political stance, cultural values and norms of the country of its origin regardless of the audience it was targeting. These differences were detected on the second-level agenda: prominence given to a story, affective tone, and frames used in a story. Thus, as a news organization attempted to adapt to a cultural setting other than its own by "domesticating" or "localizing" its agenda to its specialized target audience by addressing topics of interest to them, it still preserved its cultural and ideological identity through the attributes of these topics. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]