Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068485
Uwais Suliman, Terence L. van Zyl, A. Paskaramoorthy
Cryptocurrencies are peer-to-peer digital assets monitored and organised by a blockchain network. Price prediction has been a significant focus point with various machine learning algorithms, especially concerning cryptocurrency. This work addresses the challenge faced by traders of short-term profit maximisation. The study presents a deep reinforcement learning algorithm to trade in cryptocurrency markets, Duelling DQN. The environment has been designed to simulate actual trading behaviour, observing historical price movements and taking action on real-time prices. The proposed algorithm was tested with Bitcoin, Ethereum, and Litecoin. The respective portfolio returns are used as a metric to measure the algorithm's performance against the buy-and-hold benchmark, with the buy-and-hold outperforming the results produced by the Duelling DQN agent.
{"title":"Cryptocurrency Trading Agent Using Deep Reinforcement Learning","authors":"Uwais Suliman, Terence L. van Zyl, A. Paskaramoorthy","doi":"10.1109/ISCMI56532.2022.10068485","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068485","url":null,"abstract":"Cryptocurrencies are peer-to-peer digital assets monitored and organised by a blockchain network. Price prediction has been a significant focus point with various machine learning algorithms, especially concerning cryptocurrency. This work addresses the challenge faced by traders of short-term profit maximisation. The study presents a deep reinforcement learning algorithm to trade in cryptocurrency markets, Duelling DQN. The environment has been designed to simulate actual trading behaviour, observing historical price movements and taking action on real-time prices. The proposed algorithm was tested with Bitcoin, Ethereum, and Litecoin. The respective portfolio returns are used as a metric to measure the algorithm's performance against the buy-and-hold benchmark, with the buy-and-hold outperforming the results produced by the Duelling DQN agent.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115109828","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068440
Anand Matheven, B. V. D. Kumar
The rise of social media has brought the rise of fake news and this fake news comes with negative consequences. With fake news being such a huge issue, efforts should be made to identify any forms of fake news however it is not so simple. Manually identifying fake news can be extremely subjective as determining the accuracy of the information in a story is complex and difficult to perform, even for experts. On the other hand, an automated solution would require a good understanding of NLP which is also complex and may have difficulties producing an accurate output. Therefore, the main problem focused on this project is the viability of developing a system that can effectively and accurately detect and identify fake news. Finding a solution would be a significant benefit to the media industry, particularly the social media industry as this is where a large proportion of fake news is published and spread. In order to find a solution to this problem, this project proposed the development of a fake news identification system using deep learning and natural language processing. The system was developed using a Word2vec model combined with a Long Short-Term Memory model in order to showcase the compatibility of the two models in a whole system. This system was trained and tested using two different dataset collections that each consisted of one real news dataset and one fake news dataset. Furthermore, three independent variables were chosen which were the number of training cycles, data diversity and vector size to analyze the relationship between these variables and the accuracy levels of the system. It was found that these three variables did have a significant effect on the accuracy of the system. From this, the system was then trained and tested with the optimal variables and was able to achieve the minimum expected accuracy level of 90%. The achieving of this accuracy levels confirms the compatibility of the LSTM and Word2vec model and their capability to be synergized into a single system that is able to identify fake news with a high level of accuracy.
{"title":"Fake News Detection Using Deep Learning and Natural Language Processing","authors":"Anand Matheven, B. V. D. Kumar","doi":"10.1109/ISCMI56532.2022.10068440","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068440","url":null,"abstract":"The rise of social media has brought the rise of fake news and this fake news comes with negative consequences. With fake news being such a huge issue, efforts should be made to identify any forms of fake news however it is not so simple. Manually identifying fake news can be extremely subjective as determining the accuracy of the information in a story is complex and difficult to perform, even for experts. On the other hand, an automated solution would require a good understanding of NLP which is also complex and may have difficulties producing an accurate output. Therefore, the main problem focused on this project is the viability of developing a system that can effectively and accurately detect and identify fake news. Finding a solution would be a significant benefit to the media industry, particularly the social media industry as this is where a large proportion of fake news is published and spread. In order to find a solution to this problem, this project proposed the development of a fake news identification system using deep learning and natural language processing. The system was developed using a Word2vec model combined with a Long Short-Term Memory model in order to showcase the compatibility of the two models in a whole system. This system was trained and tested using two different dataset collections that each consisted of one real news dataset and one fake news dataset. Furthermore, three independent variables were chosen which were the number of training cycles, data diversity and vector size to analyze the relationship between these variables and the accuracy levels of the system. It was found that these three variables did have a significant effect on the accuracy of the system. From this, the system was then trained and tested with the optimal variables and was able to achieve the minimum expected accuracy level of 90%. The achieving of this accuracy levels confirms the compatibility of the LSTM and Word2vec model and their capability to be synergized into a single system that is able to identify fake news with a high level of accuracy.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114497320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068464
Alptekin Vardar, Li Zhang, Susu Hu, Saiyam Bherulal Jain, Shaown Mojumder, N. Laleni, A. Shrivastava, S. De, T. Kämpfe
Edge computing is rapidly becoming the defacto method for AI applications. However, the latency area and energy continue to be the main bottlenecks. To solve this problem, a hardware-aware approach has to be adopted. Quantizing the activations vastly reduces the number of Multiply-Accumulate (MAC) operations, resulting in with better latency and energy consumption while quantizing the weights decreases both memory footprint and the number of MAC operations, also helping with area reduction. In this paper, it is demonstrated that adapting an intra-layer mixed quantization training technique for both weights and activations, concerning layer sensitivities, in a Resnet-20 architecture with CIFAR-10 data set, a memory reduction of 73% can be achieved compared to even its all 8bits counterpart while sacrificing only around 2.3% accuracy. Moreover, it is demonstrated that, depending on the needs of the application, the balance between accuracy and resource usage can easily be arranged using different mixed-quantization schemes.
{"title":"Layer Sensitivity Aware CNN Quantization for Resource Constrained Edge Devices","authors":"Alptekin Vardar, Li Zhang, Susu Hu, Saiyam Bherulal Jain, Shaown Mojumder, N. Laleni, A. Shrivastava, S. De, T. Kämpfe","doi":"10.1109/ISCMI56532.2022.10068464","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068464","url":null,"abstract":"Edge computing is rapidly becoming the defacto method for AI applications. However, the latency area and energy continue to be the main bottlenecks. To solve this problem, a hardware-aware approach has to be adopted. Quantizing the activations vastly reduces the number of Multiply-Accumulate (MAC) operations, resulting in with better latency and energy consumption while quantizing the weights decreases both memory footprint and the number of MAC operations, also helping with area reduction. In this paper, it is demonstrated that adapting an intra-layer mixed quantization training technique for both weights and activations, concerning layer sensitivities, in a Resnet-20 architecture with CIFAR-10 data set, a memory reduction of 73% can be achieved compared to even its all 8bits counterpart while sacrificing only around 2.3% accuracy. Moreover, it is demonstrated that, depending on the needs of the application, the balance between accuracy and resource usage can easily be arranged using different mixed-quantization schemes.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123932277","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068436
M. Adegboye, Aditya Karnik, W. Fung, R. Prabhu
Pipelines are often subject to leakage due to ageing, corrosion, and weld defects, and it is difficult to avoid as the sources of leakages are diverse. Several studies have demonstrated the applicability of the machine learning model for the timely prediction of pipeline leakage. However, most of these studies rely on a large training data set for training accurate models. The cost of collecting experimental data for model training is huge, while simulation data is computationally expensive and time-consuming. To tackle this problem, the present study proposes a novel data sampling optimisation method, named adaptive particle swarm optimisation (PSO) assisted surrogate model, which was used to train the machine learning models with a limited dataset and achieved good accuracy. The proposed model incorporates the population density of training data samples and model prediction fitness to determine new data samples for improved model fitting accuracy. The proposed method is applied to 3-D pipeline leakage detection and characterisation. The result shows that the predicted leak sizes and location match the actual leakage. The significance of this study is two-fold: the practical application allows for pipeline leak prediction with limited training samples and provides a general framework for computational efficiency improvement using adaptive surrogate modelling in various real-life applications.
{"title":"Pipeline Leakage Detection and Characterisation with Adaptive Surrogate Modelling Using Particle Swarm Optimisation","authors":"M. Adegboye, Aditya Karnik, W. Fung, R. Prabhu","doi":"10.1109/ISCMI56532.2022.10068436","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068436","url":null,"abstract":"Pipelines are often subject to leakage due to ageing, corrosion, and weld defects, and it is difficult to avoid as the sources of leakages are diverse. Several studies have demonstrated the applicability of the machine learning model for the timely prediction of pipeline leakage. However, most of these studies rely on a large training data set for training accurate models. The cost of collecting experimental data for model training is huge, while simulation data is computationally expensive and time-consuming. To tackle this problem, the present study proposes a novel data sampling optimisation method, named adaptive particle swarm optimisation (PSO) assisted surrogate model, which was used to train the machine learning models with a limited dataset and achieved good accuracy. The proposed model incorporates the population density of training data samples and model prediction fitness to determine new data samples for improved model fitting accuracy. The proposed method is applied to 3-D pipeline leakage detection and characterisation. The result shows that the predicted leak sizes and location match the actual leakage. The significance of this study is two-fold: the practical application allows for pipeline leak prediction with limited training samples and provides a general framework for computational efficiency improvement using adaptive surrogate modelling in various real-life applications.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121579236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068448
Jing-Fung Lin
In this study, Taguchi design method is used to optimize the acoustic performance of two-chamber muffler. The excellent parameter combination for high signal to noise ratio (S/N) of transmission loss (TL) is obtained by the range analysis, and influence sequence of four parameters on TL is determined. TL is evaluated by a COMSOL software based on the finite element method. Further, by the modification on the radius of hole in the baffle, a revised parameter combination for better S/N is found. Finally, the stepwise regression method is used to decide a statistically significant model with a high correlation coefficient. A potential muffler is obtained by the use of genetic algorithm and has high S/N ratio of 27.097 and average value of 33.21 dB for TL in a frequency range from 10 Hz to 1400 Hz.
{"title":"Modeling and Optimization of Two-Chamber Muffler by Genetic Algorithm","authors":"Jing-Fung Lin","doi":"10.1109/ISCMI56532.2022.10068448","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068448","url":null,"abstract":"In this study, Taguchi design method is used to optimize the acoustic performance of two-chamber muffler. The excellent parameter combination for high signal to noise ratio (S/N) of transmission loss (TL) is obtained by the range analysis, and influence sequence of four parameters on TL is determined. TL is evaluated by a COMSOL software based on the finite element method. Further, by the modification on the radius of hole in the baffle, a revised parameter combination for better S/N is found. Finally, the stepwise regression method is used to decide a statistically significant model with a high correlation coefficient. A potential muffler is obtained by the use of genetic algorithm and has high S/N ratio of 27.097 and average value of 33.21 dB for TL in a frequency range from 10 Hz to 1400 Hz.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"132 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114875669","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068453
Yi-Hsuan Ting, Yi-ming Chen, Li-Kai Chen
With the improvement of computer computing speed, many researches use deep learning for Android malware detection. In addition to malware detection, malware family classification will help malware researchers understand the behavior of the malware families to optimize detection and prevent However, the new malware family has few samples, which lead to bad classification results. GAN-based method can improve the classification results, but minor data will still lead to the unstable quality of the data generated by the deep learning augmentation method, which will limit the improvement of classification results. In the study, we will propose a hybrid augmentation method, first extracting malware features and converting them into RGB images, and then the minor families will augment by the gaussian noise augmentation method, and then combined with the deep convolutional generative adversarial network (DCGAN) which have better effect on image augmentation, and finally input to CNN for family classification. The experimental results show that using the hybrid augmentation method proposed in the study, compared to no augmentation and augmentation with only using the deep convolutional generative adversarial network, the F1-Score increased between 7%~34% and 2%~7%.
{"title":"Enhancing Classification Performance for Android Small Sample Malicious Families Using Hybrid RGB Image Augmentation Method","authors":"Yi-Hsuan Ting, Yi-ming Chen, Li-Kai Chen","doi":"10.1109/ISCMI56532.2022.10068453","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068453","url":null,"abstract":"With the improvement of computer computing speed, many researches use deep learning for Android malware detection. In addition to malware detection, malware family classification will help malware researchers understand the behavior of the malware families to optimize detection and prevent However, the new malware family has few samples, which lead to bad classification results. GAN-based method can improve the classification results, but minor data will still lead to the unstable quality of the data generated by the deep learning augmentation method, which will limit the improvement of classification results. In the study, we will propose a hybrid augmentation method, first extracting malware features and converting them into RGB images, and then the minor families will augment by the gaussian noise augmentation method, and then combined with the deep convolutional generative adversarial network (DCGAN) which have better effect on image augmentation, and finally input to CNN for family classification. The experimental results show that using the hybrid augmentation method proposed in the study, compared to no augmentation and augmentation with only using the deep convolutional generative adversarial network, the F1-Score increased between 7%~34% and 2%~7%.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134339489","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068471
Doaa B. Ebaid, A. El-Zoghabi, Magda M. Madbouly
Image-text(caption) matching has become a regular evaluation of joint-embedding models that combine vision and language. This task comprises ranking the data of one modality (images) based on a Text query (Image Retrieval), and ranking texts by relevance for an image query (Text Retrieval). The current joint embedding approaches use symmetric similarity measurement, due to that order embedding is not taken in consideration. In addition to that, in image-text matching, the used losses ignore the intra similarity in a certain modality that explores the relation between the candidates in the same modality. In spite of, the important role of intra information in the embedding. In this paper, we proposed a hybrid joint embedding approach that combines between distance preserving which based on symmetric distance and order preserving that based on asymmetric distance to improve image-text matching. In addition to that we propose an intra loss function to enrich the embedding with intra-modality information. We evaluate our embedding approach on the baseline model on Flickr30K dataset. The proposed loss shows a significant enhancement in matching task.
{"title":"Hybrid Joint Embedding with Intra-Modality Loss for Image-Text Matching","authors":"Doaa B. Ebaid, A. El-Zoghabi, Magda M. Madbouly","doi":"10.1109/ISCMI56532.2022.10068471","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068471","url":null,"abstract":"Image-text(caption) matching has become a regular evaluation of joint-embedding models that combine vision and language. This task comprises ranking the data of one modality (images) based on a Text query (Image Retrieval), and ranking texts by relevance for an image query (Text Retrieval). The current joint embedding approaches use symmetric similarity measurement, due to that order embedding is not taken in consideration. In addition to that, in image-text matching, the used losses ignore the intra similarity in a certain modality that explores the relation between the candidates in the same modality. In spite of, the important role of intra information in the embedding. In this paper, we proposed a hybrid joint embedding approach that combines between distance preserving which based on symmetric distance and order preserving that based on asymmetric distance to improve image-text matching. In addition to that we propose an intra loss function to enrich the embedding with intra-modality information. We evaluate our embedding approach on the baseline model on Flickr30K dataset. The proposed loss shows a significant enhancement in matching task.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"37 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123470191","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The adoption of Artificial Intelligence (AI) is now widespread in Information Technology (IT) support. A particular area of interest is in the automation of IT incident management (i.e., the handling of an unusual event that hampers the quality of IT services in the most optimized manner). In this paper, we propose a framework using state-of-art algorithms to classify and predict the severity of such incidents (commonly labeled as High, Medium, and Low severity). We argue that the proposed framework would accelerate the process of handling IT incidents with improved accuracy. The experimentation was performed on the IT Service Management (ITSM) dataset containing 500,000 real-time incident descriptions with their encoded labels (Dataset 1) from a reputable IT firm. Our results showed that the Transformer models outperformed machine learning (ML) and other deep learning (DL) models with a 98% AUC score to predict the three severity classes. We tested our framework with an open-access dataset (Dataset 2) to further validate our findings. Our framework produced a 44% improvement in AUC score compared to the existing benchmark approaches. The results show the plausibility of AI algorithms in automating the prioritization of incident processing in large IT systems.
{"title":"Multiple Severity-Level Classifications for IT Incident Risk Prediction","authors":"Salman Ahmed, Muskaan Singh, Brendan Doherty, E. Ramlan, Kathryn Harkin, Damien Coyle","doi":"10.1109/ISCMI56532.2022.10068477","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068477","url":null,"abstract":"The adoption of Artificial Intelligence (AI) is now widespread in Information Technology (IT) support. A particular area of interest is in the automation of IT incident management (i.e., the handling of an unusual event that hampers the quality of IT services in the most optimized manner). In this paper, we propose a framework using state-of-art algorithms to classify and predict the severity of such incidents (commonly labeled as High, Medium, and Low severity). We argue that the proposed framework would accelerate the process of handling IT incidents with improved accuracy. The experimentation was performed on the IT Service Management (ITSM) dataset containing 500,000 real-time incident descriptions with their encoded labels (Dataset 1) from a reputable IT firm. Our results showed that the Transformer models outperformed machine learning (ML) and other deep learning (DL) models with a 98% AUC score to predict the three severity classes. We tested our framework with an open-access dataset (Dataset 2) to further validate our findings. Our framework produced a 44% improvement in AUC score compared to the existing benchmark approaches. The results show the plausibility of AI algorithms in automating the prioritization of incident processing in large IT systems.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130045833","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068444
K. Moloi, H. Langa
The energy supply framework has transformed over the recent time from tradition to a more flexible energy topology. This has resulted in the introduction of using distributed generation technologies to improve the sustainability of energy supply. However, the introduction of distributed generation for grid-integration has technical disadvantages, such as increase in power and voltage losses. In this paper, a technique based on discrete wavelet transform (DWT) and the genetic algorithm (GA) is proposed to determine the optimal location and size of the DGs to be connected into the grid for power loss minimisation and voltage improvement. The technique is tested using the 33 and 69 IEEE test bus system. The results obtained show that the power losses are significantly reduced with an increase in the voltage profile.
{"title":"Towards Determining the Optimal Application of Distributed Generation for Grid Integration","authors":"K. Moloi, H. Langa","doi":"10.1109/ISCMI56532.2022.10068444","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068444","url":null,"abstract":"The energy supply framework has transformed over the recent time from tradition to a more flexible energy topology. This has resulted in the introduction of using distributed generation technologies to improve the sustainability of energy supply. However, the introduction of distributed generation for grid-integration has technical disadvantages, such as increase in power and voltage losses. In this paper, a technique based on discrete wavelet transform (DWT) and the genetic algorithm (GA) is proposed to determine the optimal location and size of the DGs to be connected into the grid for power loss minimisation and voltage improvement. The technique is tested using the 33 and 69 IEEE test bus system. The results obtained show that the power losses are significantly reduced with an increase in the voltage profile.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130594244","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068468
Raha Soleymanzadeh, R. Kashef
Various research studies have been recently introduced in developing generative models, especially in computer vision and image classification. These models are inspired by a generator and discriminator network architecture in a min-max optimization game called Generative Adversarial Networks (GANs). However, GANs-based models suffer from training instability, which means high oscillations during the training, which provides inaccurate results. There are various causes beyond the instability behaviours, such as the adopted generator architecture, loss function, and distance metrics. In this paper, we focus on the impact of the generator architectures and the loss functions on the GANs training. We aim to provide a comparative assessment of various architectures focusing on ensemble and hybrid models and loss functions such as Focal loss, Binary Cross-Entropy and Mean Squared loss function. Experimental results on NSL-KDD and UNSW-NB15 datasets show that the ensemble models are more stable in terms of training and have higher intrusion detection rates. Additionally, the focal loss can improve the performance of detection minority classes. Using Mean squared loss improved the detection rate for discriminator, however with the Binary Cross entropy loss function, the deep features representation is improved and there is more stability in trends for all architectures.
{"title":"The Analysis of the Generator Architectures and Loss Functions in Improving the Stability of GANs Training towards Efficient Intrusion Detection","authors":"Raha Soleymanzadeh, R. Kashef","doi":"10.1109/ISCMI56532.2022.10068468","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068468","url":null,"abstract":"Various research studies have been recently introduced in developing generative models, especially in computer vision and image classification. These models are inspired by a generator and discriminator network architecture in a min-max optimization game called Generative Adversarial Networks (GANs). However, GANs-based models suffer from training instability, which means high oscillations during the training, which provides inaccurate results. There are various causes beyond the instability behaviours, such as the adopted generator architecture, loss function, and distance metrics. In this paper, we focus on the impact of the generator architectures and the loss functions on the GANs training. We aim to provide a comparative assessment of various architectures focusing on ensemble and hybrid models and loss functions such as Focal loss, Binary Cross-Entropy and Mean Squared loss function. Experimental results on NSL-KDD and UNSW-NB15 datasets show that the ensemble models are more stable in terms of training and have higher intrusion detection rates. Additionally, the focal loss can improve the performance of detection minority classes. Using Mean squared loss improved the detection rate for discriminator, however with the Binary Cross entropy loss function, the deep features representation is improved and there is more stability in trends for all architectures.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"162 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126807916","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：