Search

Automatic image description, also known as image captioning, aims to describe the elements included in an image and their relationships. This task involves two research fields: computer vision and natural language processing; thus, it has received much attention in computer science. In this review paper, we follow the Kitchenham review methodology to present the most relevant approaches to image description methodologies based on deep learning. We focused on works using convolutional neural networks (CNN) to extract the characteristics of images and recurrent neural networks (RNN) for automatic sentence generation. As a result, 53 research articles using the encoder-decoder approach were selected, focusing only on supervised learning. The main contributions of this systematic review are: (i) to describe the most relevant image description papers implementing an encoder-decoder approach from 2014 to 2022 and (ii) to determine the main architectures, datasets, and metrics that have been applied to image description. [ABSTRACT FROM AUTHOR]

Organizations and individuals worldwide are becoming increasingly vulnerable to cyberattacks as phishing continues to grow and the number of phishing websites grows. As a result, improved cyber defense necessitates more effective phishing detection (PD). In this paper, we introduce a novel method for detecting phishing sites with high accuracy. Our approach utilizes a Convolution Neural Network (CNN)-based model for precise classification that effectively distinguishes legitimate websites from phishing websites. We evaluate the performance of our model on the PhishTank dataset, which is a widely used dataset for detecting phishing websites based solely on Uniform Resource Locators (URL) features. Our approach presents a unique contribution to the field of phishing detection by achieving high accuracy rates and outperforming previous state-of-the-art models. Experiment results revealed that our proposed method performs well in terms of accuracy and its false-positive rate. We created a real data set by crawling 10,000 phishing URLs from PhishTank and 10,000 legitimate websites and then ran experiments using standard evaluation metrics on the data sets. This approach is founded on integrated and deep learning (DL). The CNN-based model can distinguish phishing websites from legitimate websites with a high degree of accuracy. When binary-categorical loss and the Adam optimizer are used, the accuracy of the k-nearest neighbors (KNN), Natural Language Processing (NLP), Recurrent Neural Network (RNN), and Random Forest (RF) models is 87%, 97.98%, 97.4% and 94.26%, respectively, in contrast to previous publications. Our model outperformed previous works due to several factors, including the use of more layers and larger training sizes, and the extraction of additional features from the PhishTank dataset. Specifically, our proposed model comprises seven layers, starting with the input layer and progressing to the seventh, which incorporates a layer with pooling, convolutional, linear 1 and 2, and linear six layers as the output layers. These design choices contribute to the high accuracy of our model, which achieved a 98.77% accuracy rate. [ABSTRACT FROM AUTHOR]

Deep learning technology has been extensively studied for its potential in music, notably for creative music generation research. Traditional music generation approaches based on recurrent neural networks cannot provide satisfactory long-distance dependencies. These approaches are typically designed for specific tasks, such as melody and chord generation, and cannot generate diverse music simultaneously. Pre-training is used in natural language processing to accomplish various tasks and overcome the limitation of long-distance dependencies. However, pre-training is not yet widely used in automatic music generation. Because of the differences in the attributes of language and music, traditional pre-trained models utilized in language modeling cannot be directly applied to music fields. This paper proposes a pre-trained model, MRBERT, for multitask-based music generation to learn melody and rhythm representation. The pre-trained model can be applied to music generation applications such as web-based music composers that includes the functions of melody and rhythm generation, modification, completion, and chord matching after being fine-tuned. The results of ablation experiments performed on the proposed model revealed that under the evaluation metrics of HITS@k, the pre-trained MRBERT considerably improved the performance of the generation tasks by 0.09–13.10% and 0.02–7.37%, compared to the usage of RNNs and the original BERT, respectively. [ABSTRACT FROM AUTHOR]

Understanding actions in videos remains a significant challenge in computer vision, which has been the subject of several pieces of research in the last decades. Convolutional neural networks (CNN) are a significant component of this topic and play a crucial role in the renown of Deep Learning. Inspired by the human vision system, CNN has been applied to visual data exploitation and has solved various challenges in various computer vision tasks and video/image analysis, including action recognition (AR). However, not long ago, along with the achievement of the transformer in natural language processing (NLP), it began to set new trends in vision tasks, which has created a discussion around whether the Vision Transformer models (ViT) will replace CNN in action recognition in video clips. This paper conducts this trending topic in detail, the study of CNN and Transformer for Action Recognition separately and a comparative study of the accuracy-complexity trade-off. Finally, based on the performance analysis's outcome, the question of whether CNN or Vision Transformers will win the race will be discussed. [ABSTRACT FROM AUTHOR]

Recurrent neural networks (RNNs) are widely used to process sequence-related tasks such as natural language processing. Edge cloud computing systems are in an asymmetric structure, where task managers allocate tasks to the asymmetric edge and cloud computing systems based on computation requirements. In such a computing system, cloud servers have no energy limitations, since they have unlimited energy resources. Edge computing systems, however, are resource-constrained, and the energy consumption is thus expensive, which requires an energy-efficient method for RNN job processing. In this paper, we propose a low-overhead, energy-aware runtime manager to process tasks in edge cloud computing. The RNN task latency is defined as the quality of service (QoS) requirement. Based on the QoS requirements, the runtime manager dynamically assigns RNN inference tasks to edge and cloud computing systems and performs energy optimization on edge systems using dynamic voltage and frequency scaling (DVFS) techniques. Experimental results on a real edge cloud system indicate that in edge systems, our method can reduce the energy up to 45% compared with the state-of-the-art approach. [ABSTRACT FROM AUTHOR]

Amongst energy-related CO 2 emissions, electricity is the largest single contributor, and with the proliferation of electric vehicles and other developments, energy use is expected to increase. Load forecasting is essential for combating these issues as it balances demand and production and contributes to energy management. Current state-of-the-art solutions such as recurrent neural networks (RNNs) and sequence-to-sequence algorithms (Seq2Seq) are highly accurate, but most studies examine them on a single data stream. On the other hand, in natural language processing (NLP), transformer architecture has become the dominant technique, outperforming RNN and Seq2Seq algorithms while also allowing parallelization. Consequently, this paper proposes a transformer-based architecture for load forecasting by modifying the NLP transformer workflow, adding N-space transformation, and designing a novel technique for handling contextual features. Moreover, in contrast to most load forecasting studies, we evaluate the proposed solution on different data streams under various forecasting horizons and input window lengths in order to ensure result reproducibility. Results show that the proposed approach successfully handles time series with contextual data and outperforms the state-of-the-art Seq2Seq models. [ABSTRACT FROM AUTHOR]

Recurrent neural networks (RNNs) are powerful tools for learning information from temporal sequences. Designing an optimum deep RNN is difficult due to configuration and training issues, such as vanishing and exploding gradients. In this paper, a novel metaheuristic optimisation approach is proposed for training deep RNNs for the sentiment classification task. The approach employs an enhanced Ternary Bees Algorithm (BA-3+), which operates for large dataset classification problems by considering only three individual solutions in each iteration. BA-3+ combines the collaborative search of three bees to find the optimal set of trainable parameters of the proposed deep recurrent learning architecture. Local learning with exploitative search utilises the greedy selection strategy. Stochastic gradient descent (SGD) learning with singular value decomposition (SVD) aims to handle vanishing and exploding gradients of the decision parameters with the stabilisation strategy of SVD. Global learning with explorative search achieves faster convergence without getting trapped at local optima to find the optimal set of trainable parameters of the proposed deep recurrent learning architecture. BA-3+ has been tested on the sentiment classification task to classify symmetric and asymmetric distribution of the datasets from different domains, including Twitter, product reviews, and movie reviews. Comparative results have been obtained for advanced deep language models and Differential Evolution (DE) and Particle Swarm Optimization (PSO) algorithms. BA-3+ converged to the global minimum faster than the DE and PSO algorithms, and it outperformed the SGD, DE, and PSO algorithms for the Turkish and English datasets. The accuracy value and F1 measure have improved at least with a 30–40% improvement than the standard SGD algorithm for all classification datasets. Accuracy rates in the RNN model trained with BA-3+ ranged from 80% to 90%, while the RNN trained with SGD was able to achieve between 50% and 60% for most datasets. The performance of the RNN model with BA-3+ has as good as for Tree-LSTMs and Recursive Neural Tensor Networks (RNTNs) language models, which achieved accuracy results of up to 90% for some datasets. The improved accuracy and convergence results show that BA-3+ is an efficient, stable algorithm for the complex classification task, and it can handle the vanishing and exploding gradients problem of deep RNNs. [ABSTRACT FROM AUTHOR]