Pub Date : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660152
A. J, E. Cambria, T. Trueman
Social media influences internet users to share their sentiments, feelings, or emotions about entities. In particular, sentiment analysis classifies a text into positive, negative, or neutral. It does not capture the state of mind of an individual like happiness, anger, and fear. Therefore, emotion detection plays an important role in user-generated content for capturing the state of mind. Moreover, researchers adopted traditional machine learning and deep learning models to capture emotions from the text. Recently, transformers-based architectures achieve better results in various natural language processing tasks. Therefore, we propose a transformer-based emotion detection system, which uses context-dependent features and a one-cycle learning rate policy for a better understanding of emotions from the text. We evaluate the proposed emotion detection model using error matrix, learning curve, precision, recall, F1-score, and their micro and macro averages. Our results indicate that the system achieves a 6 % accuracy over existing models.
{"title":"Transformer-Based Bidirectional Encoder Representations for Emotion Detection from Text","authors":"A. J, E. Cambria, T. Trueman","doi":"10.1109/SSCI50451.2021.9660152","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660152","url":null,"abstract":"Social media influences internet users to share their sentiments, feelings, or emotions about entities. In particular, sentiment analysis classifies a text into positive, negative, or neutral. It does not capture the state of mind of an individual like happiness, anger, and fear. Therefore, emotion detection plays an important role in user-generated content for capturing the state of mind. Moreover, researchers adopted traditional machine learning and deep learning models to capture emotions from the text. Recently, transformers-based architectures achieve better results in various natural language processing tasks. Therefore, we propose a transformer-based emotion detection system, which uses context-dependent features and a one-cycle learning rate policy for a better understanding of emotions from the text. We evaluate the proposed emotion detection model using error matrix, learning curve, precision, recall, F1-score, and their micro and macro averages. Our results indicate that the system achieves a 6 % accuracy over existing models.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127930673","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660056
Danilo Cavaliere, S. Senatore
Precision Agriculture (PA) and Forest Management (FM) applications require sensor-based environment monitoring to assess the vegetation status of monitored areas. Vegetation Indices (VIs), assessed from satellite-taken spectral images, depict some features (e.g., vegetation vigour, coverage, etc.) but they are not enough to describe vegetation status, hence they need to be contextualized according to the area phenology, latitude and weather for correct vegetation status interpretations. Moreover, heterogeneous data collection can cause data integration and interoperability issues. Additionally, human operators, who have to monitor multiple vast environments in time critical contexts, require brief meaningful reports about occurred situations. In this paper a knowledge-based multi-agent approach is presented to deal with environment monitoring of user-specified Regions of Interest (ROIs) and assess their vegetation status. The approach employs different types of agents to carry out various tasks, including data acquisition and knowledge storing, end-user interaction and vegetation analysis accomplishment. The end-user can request different types of analysis and pass data to the system through an agent-managed GUI, hence vegetation analysis is carried out by using a decision tree-based method to properly query the KB built on VIs and contextual data to consequently build a report about the vegetation status of the ROI. The built report includes a description of other features (soil, weather) that helps depicting the detected vegetation status. Several case studies demonstrate the functioning and efficacy of the approach.
{"title":"A multi-agent knowledge-enhanced model for decision-supporting agroforestry systems","authors":"Danilo Cavaliere, S. Senatore","doi":"10.1109/SSCI50451.2021.9660056","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660056","url":null,"abstract":"Precision Agriculture (PA) and Forest Management (FM) applications require sensor-based environment monitoring to assess the vegetation status of monitored areas. Vegetation Indices (VIs), assessed from satellite-taken spectral images, depict some features (e.g., vegetation vigour, coverage, etc.) but they are not enough to describe vegetation status, hence they need to be contextualized according to the area phenology, latitude and weather for correct vegetation status interpretations. Moreover, heterogeneous data collection can cause data integration and interoperability issues. Additionally, human operators, who have to monitor multiple vast environments in time critical contexts, require brief meaningful reports about occurred situations. In this paper a knowledge-based multi-agent approach is presented to deal with environment monitoring of user-specified Regions of Interest (ROIs) and assess their vegetation status. The approach employs different types of agents to carry out various tasks, including data acquisition and knowledge storing, end-user interaction and vegetation analysis accomplishment. The end-user can request different types of analysis and pass data to the system through an agent-managed GUI, hence vegetation analysis is carried out by using a decision tree-based method to properly query the KB built on VIs and contextual data to consequently build a report about the vegetation status of the ROI. The built report includes a description of other features (soil, weather) that helps depicting the detected vegetation status. Several case studies demonstrate the functioning and efficacy of the approach.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"115 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133895599","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660001
Stephen Friess, P. Tiňo, S. Menzel, B. Sendhoff, Xin Yao
Domain-dependent expertise knowledge and high-level abstractions to arbitrate between different problem domains can be considered to be essential components of how human problem-solvers build experience and reuse it over the course of their lifetime. However, replicating it from an algorithmic point of view is a less trivial endeavor. Existing knowledge transfer methods in optimization largely fail to provide more specific guidance on specifying the similarity of different optimization problems and the nature of complementary experiences formed on them. A more rigorously grounded approach can be found alternatively in metalearning. This notion neglects any hurdles on characterizing problem similarity in favor of focusing instead on methodology to form domain-dependent inductive biases and mechanisms to arbitrate between them. In principle, we proposed within our previous research methods for constructing inductive biases and predict these from procedural optimization data. However, while we obtained effective methodology, it does not allow the joint construction of predictive components and biases in a cohesive manner. We therefore show in our following study, that improved configurations can be derived for the CMA-ES algorithm which can serve as inductive biases, and that predictors can be trained to recall them. Particularly noteworthy, this scenario allows the construction of predictive component and bias iteratively in a joint manner. We demonstrate the efficacy of this approach in a shape optimization scenario, in which the inductive bias is predicted through an operator configuration in a problem-specific manner during run-time.
{"title":"Predicting CMA-ES Operators as Inductive Biases for Shape Optimization Problems","authors":"Stephen Friess, P. Tiňo, S. Menzel, B. Sendhoff, Xin Yao","doi":"10.1109/SSCI50451.2021.9660001","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660001","url":null,"abstract":"Domain-dependent expertise knowledge and high-level abstractions to arbitrate between different problem domains can be considered to be essential components of how human problem-solvers build experience and reuse it over the course of their lifetime. However, replicating it from an algorithmic point of view is a less trivial endeavor. Existing knowledge transfer methods in optimization largely fail to provide more specific guidance on specifying the similarity of different optimization problems and the nature of complementary experiences formed on them. A more rigorously grounded approach can be found alternatively in metalearning. This notion neglects any hurdles on characterizing problem similarity in favor of focusing instead on methodology to form domain-dependent inductive biases and mechanisms to arbitrate between them. In principle, we proposed within our previous research methods for constructing inductive biases and predict these from procedural optimization data. However, while we obtained effective methodology, it does not allow the joint construction of predictive components and biases in a cohesive manner. We therefore show in our following study, that improved configurations can be derived for the CMA-ES algorithm which can serve as inductive biases, and that predictors can be trained to recall them. Particularly noteworthy, this scenario allows the construction of predictive component and bias iteratively in a joint manner. We demonstrate the efficacy of this approach in a shape optimization scenario, in which the inductive bias is predicted through an operator configuration in a problem-specific manner during run-time.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131841804","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9659920
Kanwal Jahan, Jeethesh Pai Umesh, Michael Roth
This paper introduces a novel application of anomaly detection on the rail lines using deep learning methods on camera data. We propose a two-fold approach for identifying irregularities like coal, dirt, and obstacles on the rail tracks. In the first stage, a binary semantic segmentation is performed to extract only the rails from the background. In the second stage, we deploy our proposed autoencoder utilizing the self-supervised learning techniques to address the unavailability of labelled anomalies. The extracted rails from stage one are divided into multiple patches and are fed to the autoencoder, which is trained to reconstruct the non-anomalous data only. Hence, during the inference, the regeneration of images with any abnormalities produces a larger reconstruction error. Applying a predefined threshold to the reconstruction errors can detect an anomaly on a rail track. Stage one, rail extracting network achieves a high value of 52.78% mean Intersection over Union (mIoU). The second stage autoencoder network converges well on the training data. Finally, we evaluate our two-fold approach on real scenario test images, no false positives or false negatives were found in the the detected anomalies on the rail tracks.
{"title":"Anomaly Detection on the Rail Lines Using Semantic Segmentation and Self-supervised Learning","authors":"Kanwal Jahan, Jeethesh Pai Umesh, Michael Roth","doi":"10.1109/SSCI50451.2021.9659920","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659920","url":null,"abstract":"This paper introduces a novel application of anomaly detection on the rail lines using deep learning methods on camera data. We propose a two-fold approach for identifying irregularities like coal, dirt, and obstacles on the rail tracks. In the first stage, a binary semantic segmentation is performed to extract only the rails from the background. In the second stage, we deploy our proposed autoencoder utilizing the self-supervised learning techniques to address the unavailability of labelled anomalies. The extracted rails from stage one are divided into multiple patches and are fed to the autoencoder, which is trained to reconstruct the non-anomalous data only. Hence, during the inference, the regeneration of images with any abnormalities produces a larger reconstruction error. Applying a predefined threshold to the reconstruction errors can detect an anomaly on a rail track. Stage one, rail extracting network achieves a high value of 52.78% mean Intersection over Union (mIoU). The second stage autoencoder network converges well on the training data. Finally, we evaluate our two-fold approach on real scenario test images, no false positives or false negatives were found in the the detected anomalies on the rail tracks.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131889272","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9659992
A. Thavaneswaran, R. Thulasiram, Md. Erfanul Hoque, S. S. Appadoo
Uncertainty in supply chain leads to what is known as bullwhip effect (BE), which causes multiple inefficiencies such as higher costs of production (of more than what is needed), wastage and logistics. Though there are many studies reported in the literature, the impact of the quality of dynamic forecasts on the BE has not received sufficient coverage. In this paper, a fuzzy data-driven weighted moving average (DDWMA) forecasts of the future demand strategy is proposed for supply chain. Also, data-driven random weighted volatility forecasting model is used to study the fuzzy extended Bollinger bands forecasts of the demand. The main reason of using the fuzzy approach is to provide α-cuts for DDWMA demand forecasts as well as extended Bollinger bands forecasts. The proposed fuzzy extended Bollinger bands forecast is a two steps procedure as it uses optimal weights for both the demand forecasts as well as the volatility forecasts of the demand process. In particular, a novel dynamic fuzzy forecasting algorithm of the demand is proposed which bypasses complexities associated with traditional forecasting steps of fitting any time series model. The proposed data-driven fuzzy forecasting approach focuses on defining a dynamic fuzzy forecasting intervals of the demand as well as the volatility of the demand in supply chain. The performance of proposed approaches is evaluated through numerical experiments using simulated data and weekly demand data. The results show that the proposed methods perform well in terms of narrower fuzzy forecasting bands for demand as well as the volatility of the demand.
{"title":"Data-Driven Fuzzy Demand Forecasting Models for Resilient Supply Chains","authors":"A. Thavaneswaran, R. Thulasiram, Md. Erfanul Hoque, S. S. Appadoo","doi":"10.1109/SSCI50451.2021.9659992","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659992","url":null,"abstract":"Uncertainty in supply chain leads to what is known as bullwhip effect (BE), which causes multiple inefficiencies such as higher costs of production (of more than what is needed), wastage and logistics. Though there are many studies reported in the literature, the impact of the quality of dynamic forecasts on the BE has not received sufficient coverage. In this paper, a fuzzy data-driven weighted moving average (DDWMA) forecasts of the future demand strategy is proposed for supply chain. Also, data-driven random weighted volatility forecasting model is used to study the fuzzy extended Bollinger bands forecasts of the demand. The main reason of using the fuzzy approach is to provide α-cuts for DDWMA demand forecasts as well as extended Bollinger bands forecasts. The proposed fuzzy extended Bollinger bands forecast is a two steps procedure as it uses optimal weights for both the demand forecasts as well as the volatility forecasts of the demand process. In particular, a novel dynamic fuzzy forecasting algorithm of the demand is proposed which bypasses complexities associated with traditional forecasting steps of fitting any time series model. The proposed data-driven fuzzy forecasting approach focuses on defining a dynamic fuzzy forecasting intervals of the demand as well as the volatility of the demand in supply chain. The performance of proposed approaches is evaluated through numerical experiments using simulated data and weekly demand data. The results show that the proposed methods perform well in terms of narrower fuzzy forecasting bands for demand as well as the volatility of the demand.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132197570","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660133
Iron Tessaro, R. Z. Freire, V. Mariani, L. Coelho
One of the applications of data-driven methods in the industry is the creation of real-time, embedded measurements, whether to monitor or replace sensor signals. As the number of embedded systems in products raises over time, the energy efficiency of such systems must be considered in the design. The time (processor) efficiency of the embedded software is directly related to the energy efficiency of the embedded system. Therefore, when considering some embedded software solutions, such as data-driven methods, time efficiency must be taken into account to improve energy efficiency. In this work, the energy efficiency of three data-driven methods: the Sparse Identification of Nonlinear Dynamics (SINDy), the Extreme Learning Machine (ELM), and the Random-Vector Functional Link (RVFL) network were assessed by using the creation of a real-time in-cylinder pressure sensor for diesel engines as a task. The three methods were kept with equivalent performances, whereas their relative execution time was tested and classified by their statistical rankings. Additionally, the space (memory) efficiency of the methods was assessed. The contribution of this work is to provide a guide to choose the best data-driven method to be used in an embedded system in terms of efficiency.
{"title":"Space and Time Efficiency Analysis of Data-Driven Methods Applied to Embedded Systems","authors":"Iron Tessaro, R. Z. Freire, V. Mariani, L. Coelho","doi":"10.1109/SSCI50451.2021.9660133","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660133","url":null,"abstract":"One of the applications of data-driven methods in the industry is the creation of real-time, embedded measurements, whether to monitor or replace sensor signals. As the number of embedded systems in products raises over time, the energy efficiency of such systems must be considered in the design. The time (processor) efficiency of the embedded software is directly related to the energy efficiency of the embedded system. Therefore, when considering some embedded software solutions, such as data-driven methods, time efficiency must be taken into account to improve energy efficiency. In this work, the energy efficiency of three data-driven methods: the Sparse Identification of Nonlinear Dynamics (SINDy), the Extreme Learning Machine (ELM), and the Random-Vector Functional Link (RVFL) network were assessed by using the creation of a real-time in-cylinder pressure sensor for diesel engines as a task. The three methods were kept with equivalent performances, whereas their relative execution time was tested and classified by their statistical rankings. Additionally, the space (memory) efficiency of the methods was assessed. The contribution of this work is to provide a guide to choose the best data-driven method to be used in an embedded system in terms of efficiency.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"34 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133801415","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660090
Tsung-Su Yeh, T. Chiang
A hybrid flow shop is a kind of flow shop where multiple machines are available at some stages. This paper addresses the hybrid flow shop scheduling problem (HFSP) with identical parallel machines. We propose an algorithm based on the framework of Leaders and Followers (LaF), a recent metaheuristic that searches by two populations. We apply iterated greedy (IG) to the leader population for exploitation and genetic algorithm (GA) to the follower population for exploration. Investigations on the parameter setting and technical details of the algorithm are made by experiments using 240 public problem instances. Performance comparison with two recent algorithms verifies the solution quality and computational efficiency of the proposed algorithm.
{"title":"Hybrid Flowshop Scheduling using Leaders and Followers: An Implementation with Iterated Greedy and Genetic Algorithm","authors":"Tsung-Su Yeh, T. Chiang","doi":"10.1109/SSCI50451.2021.9660090","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660090","url":null,"abstract":"A hybrid flow shop is a kind of flow shop where multiple machines are available at some stages. This paper addresses the hybrid flow shop scheduling problem (HFSP) with identical parallel machines. We propose an algorithm based on the framework of Leaders and Followers (LaF), a recent metaheuristic that searches by two populations. We apply iterated greedy (IG) to the leader population for exploitation and genetic algorithm (GA) to the follower population for exploration. Investigations on the parameter setting and technical details of the algorithm are made by experiments using 240 public problem instances. Performance comparison with two recent algorithms verifies the solution quality and computational efficiency of the proposed algorithm.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115566127","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660140
H. C. R. Oliveira, V. Shmerko, S. Yanushkevich
This paper focuses on designing a CI decision support to address rare events such as disease outbreaks in a ‘closed’ environment such as a cruise ship. We focus on a case study of the COVID-19 outbreak that happened on board the Diamond Princess cruise ship in 2020. It considers a graphical probabilistic model such as Bayesian Network. We consider this causal model to be a core of an intelligent decision support tool to help in emergency management. To prove this hypothesis, the prototype of a decision support tool was implemented and used to evaluate different scenarios. The results show that such system equipped with a reasoning engine is capable of evaluating the pandemic scenario risks, thus helping assess the impacts of certain preventive measures, and damages.
{"title":"Decision Support for Infection Outbreak Analysis: the case of the Diamond Princess cruise ship","authors":"H. C. R. Oliveira, V. Shmerko, S. Yanushkevich","doi":"10.1109/SSCI50451.2021.9660140","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660140","url":null,"abstract":"This paper focuses on designing a CI decision support to address rare events such as disease outbreaks in a ‘closed’ environment such as a cruise ship. We focus on a case study of the COVID-19 outbreak that happened on board the Diamond Princess cruise ship in 2020. It considers a graphical probabilistic model such as Bayesian Network. We consider this causal model to be a core of an intelligent decision support tool to help in emergency management. To prove this hypothesis, the prototype of a decision support tool was implemented and used to evaluate different scenarios. The results show that such system equipped with a reasoning engine is capable of evaluating the pandemic scenario risks, thus helping assess the impacts of certain preventive measures, and damages.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114615739","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9659550
Asma Sattar, D. Bacciu
Graph neural networks allow to build recommendation systems which can straightforwardly take into account relational knowledge concerning multiple types of interactions, such as user-item relationships, but also interactions between users and within items. Graph-based approaches in the literature consider such interactions to be static, independent of the surroundings. In this paper, we put forward a novel approach to graph-based item recommendation built on the foundational idea that relational knowledge is characterized by a dynamic nature of the user and its surroundings. We claim that being able to capture such dynamic knowledge allows to build richer contexts upon which more precise recommendations can be built, e.g., taking into account current location, weather conditions, and user mood. The paper provides recipes to build and integrate dynamic user and item contexts in existing item recommendation tasks. We also introduce a novel Dynamic Context-aware Graph Neural Network (DCGNN) that can effectively leverage the knowledge of surroundings to learn the context-aware recommendation behaviour of users. The empirical analysis shows how our model outperforms static state-of-the-art approaches on four movie and travel recommendation benchmarks.
{"title":"Dynamic Context in Graph Neural Networks for Item Recommendation","authors":"Asma Sattar, D. Bacciu","doi":"10.1109/SSCI50451.2021.9659550","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659550","url":null,"abstract":"Graph neural networks allow to build recommendation systems which can straightforwardly take into account relational knowledge concerning multiple types of interactions, such as user-item relationships, but also interactions between users and within items. Graph-based approaches in the literature consider such interactions to be static, independent of the surroundings. In this paper, we put forward a novel approach to graph-based item recommendation built on the foundational idea that relational knowledge is characterized by a dynamic nature of the user and its surroundings. We claim that being able to capture such dynamic knowledge allows to build richer contexts upon which more precise recommendations can be built, e.g., taking into account current location, weather conditions, and user mood. The paper provides recipes to build and integrate dynamic user and item contexts in existing item recommendation tasks. We also introduce a novel Dynamic Context-aware Graph Neural Network (DCGNN) that can effectively leverage the knowledge of surroundings to learn the context-aware recommendation behaviour of users. The empirical analysis shows how our model outperforms static state-of-the-art approaches on four movie and travel recommendation benchmarks.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114758417","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 : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660025
Mustafa Misir
The present study aims at generating heuristics for Protein Structure Prediction represented in the 2D HP model. Protein Structure Prediction is about determining the 3-dimensional form of a protein from a given amino acid sequence. The resulting structure directly relates to the functionalities of the protein. There are a wide range of algorithms to address Protein Structure Prediction as an optimization problem. Being said that there is no an ultimate algorithm that can effectively solve PSP under varying experimental settings. Hyper-heuristics can offer a solution as high-level, problem-independent search and optimization strategies. Selection Hyper-heuristics operate on given heuristic sets that directly work on the solution space. One group of Selection Hyper-heuristics focus on automatically specify the best heuristics on-the-fly. Yet, the candidate heuristics tend to be decided, preferably a domain expert. Generation Hyper-heuristics approach differently as aiming to generate such heuristics automatically. This work introduces a automated heuristic generation strategy supporting Selection Hyper-heuristics. The generation task is formulated as a selection problem, disclosing the best expected heuristic specifically f or a given problem instance. The heuristic generation process is established as a parameter configuration problem. T he corresponding system is devised by initially generating a training data alongside with a set of basic features characterizing the Protein Structure Prediction problem instances. The data is generated discretizing the parameter configuration space o f a single heuristic. The resulting data is used to predict the best configuration of a specific heuristic used in a heuristic set under Selection Hyper-heuristics. The prediction is performed separately for each instance rather than using one setting for all the instances. The empirical analysis showed that the proposed idea offers both better and robust performance on 22 PSP instances compared to the one-for-all heuristic sets. Additional analysis linked to the selection method, ALORS, revealed insights on what makes the PSP instances hard / easy while providing dis/-similarity analysis between the candidate configurations.
{"title":"Selection-based Per-Instance Heuristic Generation for Protein Structure Prediction of 2D HP Model","authors":"Mustafa Misir","doi":"10.1109/SSCI50451.2021.9660025","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660025","url":null,"abstract":"The present study aims at generating heuristics for Protein Structure Prediction represented in the 2D HP model. Protein Structure Prediction is about determining the 3-dimensional form of a protein from a given amino acid sequence. The resulting structure directly relates to the functionalities of the protein. There are a wide range of algorithms to address Protein Structure Prediction as an optimization problem. Being said that there is no an ultimate algorithm that can effectively solve PSP under varying experimental settings. Hyper-heuristics can offer a solution as high-level, problem-independent search and optimization strategies. Selection Hyper-heuristics operate on given heuristic sets that directly work on the solution space. One group of Selection Hyper-heuristics focus on automatically specify the best heuristics on-the-fly. Yet, the candidate heuristics tend to be decided, preferably a domain expert. Generation Hyper-heuristics approach differently as aiming to generate such heuristics automatically. This work introduces a automated heuristic generation strategy supporting Selection Hyper-heuristics. The generation task is formulated as a selection problem, disclosing the best expected heuristic specifically f or a given problem instance. The heuristic generation process is established as a parameter configuration problem. T he corresponding system is devised by initially generating a training data alongside with a set of basic features characterizing the Protein Structure Prediction problem instances. The data is generated discretizing the parameter configuration space o f a single heuristic. The resulting data is used to predict the best configuration of a specific heuristic used in a heuristic set under Selection Hyper-heuristics. The prediction is performed separately for each instance rather than using one setting for all the instances. The empirical analysis showed that the proposed idea offers both better and robust performance on 22 PSP instances compared to the one-for-all heuristic sets. Additional analysis linked to the selection method, ALORS, revealed insights on what makes the PSP instances hard / easy while providing dis/-similarity analysis between the candidate configurations.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115305720","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：