filter by:
Articles
Multilingual neural machine translation (MNMT) is a novel machine translation approach that benefits from large multilingual resources. However, its performance drops significantly when training with low-resource languages due to the reliance on parameter sharing and data size. In this paper, a new method is proposed to improve the performance of MNMT for a pair of languages where the target language is low-resource. The main idea of this study is to find important nodes that have parameters connected to them that negatively affect an MNMT model and then split those nodes into two sub nodes. Then, the model selects the important sub node that has an effect on the specific language pair to create a twin sub node. This twin sub node helps to strengthen the translation quality of the specific language pair without having a negative effect on other languages. The proposed method works in four steps as: 1) training an MNMT model with parameter sharing over multiple languages, 2) selecting important nodes which negatively affect the MNMT, 3) splitting important nodes into sub nodes, and 4) Twining important sub nodes. The proposed method has been evaluated using several multilingual datasets, including TED 2013, TED 2020, and BIBLE, by examining English-Persian language as a case study. The obtained results show that the proposed method yields the best results for one-to-many and many-to-many models according to the average BLEU value and semantic similarity. The results also show that the proposed method has given better results than other well-known large language models, such as ChatGPT, BING GPT4, and the Google Neural Machine Translation (GNMT) model, when applied to a low-resource language. © 2025 Elsevier B.V.
2025 29th International Computer Conference, Computer Society of Iran, CSICC 2025pp. 253-259
This study presents a method for the automatic identification of micro-cracks in photovoltaic solar modules using deep learning techniques. The main challenge in this research is the lack of labeled data and class imbalance for the detection of micro-cracks. The proposed method employs a multi-stage approach. Initially, 10% of the dataset is manually labeled to train a simple convolutional neural network model. This model is then used to generate pseudo-labels for the unlabeled data using a semi-supervised approach. The pseudo-labels are manually reviewed to increase the number of micro-crack samples in the training set. Data augmentation techniques are also applied to increase the size and diversity of the training dataset. Finally, the pre-trained ResNet-50 model is fine-tuned on the expanded labeled dataset for accurate detection of microcracks. Advanced preprocessing steps, including solar cell segmentation, cropping, and data augmentation, have been performed. The class imbalance problem is addressed through undersampling and weighted loss functions. The experimental results demonstrate the effectiveness of the proposed method, achieving an accuracy of 0.978 and an F1-score of 0.797 in the detection of micro-cracks in electroluminescence images of solar panels. This study provides insights into the use of limited labeled data for training robust deep learning models for the identification of defects in solar modules. © 2024 IEEE.
International Journal of Machine Learning and Cybernetics (18688071)(12)pp. 5509-5529
Federated semi-supervised learning (Fed-SSL) algorithms have been developed to address the challenges of decentralized data access, data confidentiality, and costly data labeling in distributed environments. Most existing Fed-SSL algorithms are based on the federated averaging approach, which utilizes an equivalent model on all machines and replaces local models during the learning process. However, these algorithms suffer from significant communication overhead when transferring parameters of local models. In contrast, knowledge distillation-based Fed-SSL algorithms reduce communication costs by only transferring the output of local models on shared data between machines. However, these algorithms assume that all local data on the machines are labeled, and that there exists a large set of shared unlabeled data for training. These assumptions are not always feasible in real-world applications. In this paper, a knowledge distillation-based Fed-SSL algorithm has been presented, which does not make any assumptions about how the data is distributed among machines. Additionally, it artificially generates shared data required for the learning process. The learning process of the presented approach employs a semi-supervised GAN on local machines and has two stages. In the first stage, each machine trains its local model independently. In the second stage, each machine generates some artificial data in each step and propagates it to other machines. Each machine trains its discriminator with these data and the average output of all machines on these data. The effectiveness of this algorithm has been examined in terms of accuracy and the amount of communication among machines by using different data sets with different distributions. The evaluations reveal that, on average, the presented algorithm is 15% more accurate than state-of-the-art methods, especially in the case of non-IID data. In addition, in most cases, it yields better results than existing studies in terms of the amount of data communication among machines. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.
Iranian Journal Of Fuzzy Systems (17350654)(6)pp. 21-47
Visual cryptography is a method used to secure images by converting them into several shares that are finally stacked to recover the original image without any calculations. Most existing techniques that can encrypt gray and color images often convert them to binary or support only limited colors, which results in reducing the quality of recovered images. Pixel expansion is another problem with existing methods. Thus, a new approach is required to encrypt gray and color images with real value, without converting them to binary or limited-color images, and also without imposing any pixel expansions. Besides, generated shares should have security, and the recovered images should be of high quality. In this research, fuzzy random grids and a meta-heuristic algorithm are used for the share generation in the encryption step. Next, the decryption step uses the fuzzy OR operator to recover high-quality images. The evaluation results demonstrate the ability of the proposed solution in encrypting gray and color images without converting them to binary, and also without pixel expansion. Besides, the results show that the proposed method is secure as individual shares do not show any information from the original image. The quality of the decrypted images has also been evaluated using subjective and objective evaluation metrics, which prove the high quality of recovered images. © 2023, University of Sistan and Baluchestan. All rights reserved.
Journal of Biomedical Informatics (15320464)
Summarization is the process of compressing a text to obtain its important informative parts. In recent years, various methods have been presented to extract important parts of textual documents to present them in a summarized form. The first challenge of these methods is to detect the concepts that well convey the main topic of the text and extract sentences that better describe these essential concepts. The second challenge is the correct interpretation of the essential concepts to generate new paraphrased sentences such that they are not exactly the same as the sentences in the main text. The first challenge has been addressed by many researchers. However, the second one is still in progress. In this study, we focus on the abstractive summarization of biomedical documents. In this regard, for the first challenge, a new method is presented based on the graph generation and frequent itemset mining for generating extractive summaries by considering the concepts within the biomedical documents. Then, to address the second challenge, a transfer learning-based method is used to generate abstractive summarizations from extractive summaries. The efficiency of the proposed solution has been evaluated by conducting several experiments over BioMed Central and NLM's PubMed datasets. The obtained results show that the proposed approach admits a better interpretation of the main concepts and sentences of biomedical documents for the abstractive summarization by obtaining the overall ROUGE of 59.60%, which, on average, is 17% better than state-of-the-art summarization techniques. The source code, datasets, and results are available in GitHub1. © 2022 Elsevier Inc.
Expert Systems with Applications (9574174)
In the Authorship Attribution (AA) task, the most likely author of textual documents, such as books, papers, news, and text messages and posts are identified using statistical and computational methods. In this paper, a new computational approach is presented for identifying the most likely author of text documents. The proposed solution emphasizes lazy profile-based classification and, by using the Term Frequency-Inverse Document Frequency (TF_IDF) scheme, introduces a new measure for identifying important terms of documents. The importance of the terms is then used to calculate the similarity between an anonymous document and known documents. The proposed solution works with raw text documents and does not require any NLP tools for preprocessing, which makes it language-independent. The efficiency of the proposed solution has been evaluated by conducting several experiments on two English and Persian datasets, each of which contains six corpora with different number of authors. The obtained results demonstrate that the proposed solution outperforms state-of-the-art stylometric features, employed by seven well-known classifiers, by obtaining 0.902 accuracy for the English dataset and 0.931 accuracy for the Persian dataset. In addition, supplementary experiments have been conducted to evaluate the effects of documents’ length on the accuracy of the proposed solution, to examine the computation time of the proposed solution and competitive classifiers, and to identify the most effective stylometric features and classifiers. © 2021 Elsevier Ltd
Journal of Supercomputing (9208542)(1)pp. 597-618
SRAM-based FPGAs feature high performance and flexibility. Thus, they have found many applications in modern high-performance computing (HPC) systems. These systems suffer from the limitation of the computing resources problem for running HPC applications. Therefore, multi-FPGA systems have been emerged to alleviate such resource limitations. In this regard, efficient scheduling strategies are required to dynamically steer the execution of applications—represented as task graphs—on a set of connected FPGAs. In this paper, a heuristic-based dynamic critical path-aware scheduling technique named CPA is presented to schedule task graphs on multi-FPGA systems. The proposed technique, by considering the computation and communication capabilities of FPGAs, dynamically assigns priority to tasks in different steps in order to achieve better makespans. The proposed technique has been evaluated by conducting several experiments on real-world and three different shapes of random task graphs with different number of tasks, and its efficiency has been compared with that of three task graph scheduling approaches. The obtained results demonstrate that the proposed CPA technique outperforms well-known heuristic scheduling strategies and improves their makespan by 13.47% on average. In addition, the experiments show that the proposed technique generates the schedules in the order of milliseconds and the average of its yielded makespans is 12.05% longer than that of an optimum schedule. © 2020, Springer Science+Business Media, LLC, part of Springer Nature.
Reliability Engineering and System Safety (9518320)
SRAM-based FPGAs have found many applications in modern computer systems. In these systems, high-performance computing applications are executed as task graphs in which reliability and performance are crucial constraints. In this paper, an exact method is presented to efficiently optimize the reliability and performance of synchronous task graphs running on SRAM-based FPGAs in harsh environments. Solving this optimization problem leads to the generation of a true Pareto set of Fault Tolerance (FT) techniques. Each solution of this set determines FT techniques of the tasks and leads to specific reliability and makespan. Thus, this solution set trades off between reliability and makespan, and one of the solutions that best meets system requirements can be applied to the running tasks to optimize the reliability and performance. The proposed technique is novel as it obtains the true Pareto set of FT techniques of the whole task graph by partitioning the task graph into its segments, optimizing different segments separately, and joining the obtained solutions. This partitioning strategy leads to reduce the computation time significantly. In this paper, it is mathematically proved that the proposed partitioning strategy generates global optima from the local ones without losing any optimal solutions. The experiments show that the proposed technique improves the MTTF of real-world and random task graphs by 46.30% on average without any negative effects on the performance. Then, the efficiency of the computation time of the proposed technique is demonstrated by conducting several experiments on small-, medium-, and large-size synchronous task graphs and comparing the results with other exact and evolutionary optimization methods. Finally, supplementary experiments in dynamic environments show that the proposed technique outperforms adaptive state-of-the-art FT techniques in terms of reliability and makespan improvement. © 2020 Elsevier Ltd
Journal of Supercomputing (9208542)(9)pp. 7140-7160
In partially run-time reconfigurable (PRR) FPGAs, hardware tasks should be configured before their execution. The configuration delay imposed by the reconfiguration process increases the total execution time of the hardware tasks and task graphs. In this paper, a new technique named forefront-fetch is presented to improve the makespan of hardware task graphs running on PRR FPGAs via alleviating the adverse effects of the configuration delays. In this technique, which is applied to a sequence of task graphs, the configuration of some tasks is carried out within the execution phase of the previous task graph. This strategy leads to hide the configuration delay of the forefront-fetched tasks that as a result improves the execution time. The proposed solution modifies the schedules of the task graphs at design time to obtain a set of schedule pairs for the run-time environment. Experiments on actual and synthesized task graphs demonstrate the ability of the proposed technique in improving the makespan of hardware task graphs. The obtained results show that for a set of task graphs running on Xilinx™ Virtex-5 XUPV5LX110T FPGA, makespan is improved by 37.81% on average. Moreover, the proposed solution outperforms state-of-the-art prefetch-aware scheduling strategies by 14.78% makespan improvement. © 2020, Springer Science+Business Media, LLC, part of Springer Nature.
Reliability Engineering and System Safety (9518320)
This paper addresses the problem of hardware tasks reliability estimation in harsh environments. A novel statistical model is presented to estimate the reliability, the mean time to failure, and the number of errors of hardware tasks running on static random-access memory (SRAM)-based partially run-time reconfigurable field programmable gate arrays (FPGAs) in harsh environments by taking both single-bit upsets and multiple-cell upsets into account. The model requires some features of the hardware tasks, including their computation time, size, the percent of critical bits, and the soft error rates of k-bit events (k ≥ 1) of the environment for the reliability estimation. Such an early estimation helps the developers to assess the reliability of their designs at earlier stages and leads to reduce the development cost. The proposed model has been evaluated by conducting several experiments on actual hardware tasks over different environmental soft error rates. The obtained results, endorsed by the 95% confidence interval, reveal the high accuracy of the proposed model. When comparing this approach with a reliability model (developed by the authors in a previous work) that does not consider the occurrence of multiple-cell upsets, an overestimation of the mean time to failure of 2.88X is observable in the latter. This points to the importance of taking into account multiple events, especially in modern technologies where the miniaturization is high. © 2020 Elsevier Ltd
Reliability Engineering and System Safety (9518320)pp. 13-24
This paper presents an approach to optimize the reliability and makespan of hardware task graphs, running on FPGA-based reconfigurable computers, in space-mission computing applications with dynamic soft error rates (SERs). Thus, with rises and falls of the SER, the presented approach dynamically generates a set of solutions that apply redundancy-based fault tolerance (FT) techniques to the running tasks. The set of solutions is generated by decomposing the task graph into multiple subgraphs, applying a multi-objective optimization algorithm to the subgraphs separately, and finally combining and filtering out the obtained solutions of the subgraphs. In this regard, a heuristic has been proposed to decompose task graphs in such a way that a high coverage of the true Pareto set is attained. The experiments show that the presented approach covers 97.37% of the true Pareto set and improves the average computation time of generating the Pareto set from 6.29 h to 81.86 ms. In addition, it outperforms the NSGA-II algorithm in terms of the Pareto set coverage and computation time. Additional experiments demonstrate the advantages of the presented approach over the state-of-the-art adaptive FT techniques in dynamic environments. © 2018 Elsevier Ltd
