Pub Date : 2022-11-26DOI: 10.1109/ISCMI56532.2022.10068438
James Sanii, Wai-Yip Chan
To gain trust, machine learning (ML) models used in high stake applications such as clinical decision support need to provide explainable behaviours and outputs. To assess whether interpretable explanations can be obtained without sacrificing prediction performance, we compare using “black box” versus “glass box” models for predicting the mortality risk of patients diagnosed with pneumonia, using data in the MIMIC-III dataset. We examine five types of black box models: random forest (RF), support vector machine (SVM), gradient boosting classifier (GBC), AdaBoost (ADA), and multilayer perceptron (MLP), and three types of glassbox models: K-nearest neighbor (KNN), explainable boosting machine (EBM), and generalized additive models (GAM). When trained using 417 features, a black box RF model performs best with AUC of 0.896. With the feature set size reduced to 19, an EBM model performs the best with AUC 0.872. Both models exceed the AUC of 0.661, the best previously reported for the task. Our results suggest that ML models with inbuilt explainability may provide prediction power as attractive as black box models.
{"title":"Explainable Machine Learning Models for Pneumonia Mortality Risk Prediction Using MIMIC-III Data","authors":"James Sanii, Wai-Yip Chan","doi":"10.1109/ISCMI56532.2022.10068438","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068438","url":null,"abstract":"To gain trust, machine learning (ML) models used in high stake applications such as clinical decision support need to provide explainable behaviours and outputs. To assess whether interpretable explanations can be obtained without sacrificing prediction performance, we compare using “black box” versus “glass box” models for predicting the mortality risk of patients diagnosed with pneumonia, using data in the MIMIC-III dataset. We examine five types of black box models: random forest (RF), support vector machine (SVM), gradient boosting classifier (GBC), AdaBoost (ADA), and multilayer perceptron (MLP), and three types of glassbox models: K-nearest neighbor (KNN), explainable boosting machine (EBM), and generalized additive models (GAM). When trained using 417 features, a black box RF model performs best with AUC of 0.896. With the feature set size reduced to 19, an EBM model performs the best with AUC 0.872. Both models exceed the AUC of 0.661, the best previously reported for the task. Our results suggest that ML models with inbuilt explainability may provide prediction power as attractive as black box models.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"83 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":"130334679","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/ISCMI56532.2022.10068445
M. Khumalo, K. Prag, K. Nixon
The performance of the Quantum Approximate Optimisation Algorithm (QAOA) in solving the Quadratic Assignment Problem (QAP) is evaluated, with the Variational Quantum Eigensolver (VQE) as a benchmark. The QAP is directly revelant to numerous industry scenarios. The QAP, a Combinatorial Optimisation Problem (COP), is classified as $mathcal{NP}$ -Hard. This classification means CPU time increases exponentially as the problem size scales when solving the QAP using deterministic optimisation techniques. Therefore, this work investigates the QAOA in search of a non-deterministic optimisation technique to efficiently obtain solutions to the QAP. This research compares two warm start techniques to solve QAP instances of sizes 3 to 7. The metrics of comparison - that measure efficiency and solution quality - were introduced in previous work on this topic. For the QAOA, the impact of the p-value, a determination of circuit depth, is investigated. Of the two quantum hybrid heuristics, the VQE retrieves solutions in a shorter computational time with a smaller circuit size, which allows for solving instances with a larger problem size. Compared to the VQE, the QAOA performs better in terms of feasibility as the problem size scales. The quantum warm start method results implies that the QAOA may not maintain higher solution quality for instances larger than size 4. Still, further investigation should be conducted once quantum devices with more qubits and higher quantum volumes are available.
{"title":"Evaluating the performance of the Quantum Approximate Optimisation Algorithm to solve the Quadratic Assignment Problem","authors":"M. Khumalo, K. Prag, K. Nixon","doi":"10.1109/ISCMI56532.2022.10068445","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068445","url":null,"abstract":"The performance of the Quantum Approximate Optimisation Algorithm (QAOA) in solving the Quadratic Assignment Problem (QAP) is evaluated, with the Variational Quantum Eigensolver (VQE) as a benchmark. The QAP is directly revelant to numerous industry scenarios. The QAP, a Combinatorial Optimisation Problem (COP), is classified as $mathcal{NP}$ -Hard. This classification means CPU time increases exponentially as the problem size scales when solving the QAP using deterministic optimisation techniques. Therefore, this work investigates the QAOA in search of a non-deterministic optimisation technique to efficiently obtain solutions to the QAP. This research compares two warm start techniques to solve QAP instances of sizes 3 to 7. The metrics of comparison - that measure efficiency and solution quality - were introduced in previous work on this topic. For the QAOA, the impact of the p-value, a determination of circuit depth, is investigated. Of the two quantum hybrid heuristics, the VQE retrieves solutions in a shorter computational time with a smaller circuit size, which allows for solving instances with a larger problem size. Compared to the VQE, the QAOA performs better in terms of feasibility as the problem size scales. The quantum warm start method results implies that the QAOA may not maintain higher solution quality for instances larger than size 4. Still, further investigation should be conducted once quantum devices with more qubits and higher quantum volumes are available.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"60 3 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":"128013900","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/ISCMI56532.2022.10068465
Szilvia Jáhn-Erdös, Bence Kövári
Genetic algorithms for NP-complete problems are widespread since it is easy to obtain a solution to the problem. However, its optimality gives the real issue and is not guaranteed to be achievable. In our research, we address a special subproblem of scheduling problems, the final exam scheduling, in which special requirements restrict the state space, which often contradicts each other. The task's difficulty is the massive size of the state space. Genetic algorithm-based solutions were considered since a MILP solver could not find a solution in a reasonable time. A model was built to solve this problem using the genetic algorithm. Most of the possibilities were seen in the different mutation procedures, so we investigated them in more detail. A question for genetic algorithms is what parameters and probabilities to run the model with since the more freedom we give to the run, the larger the runtime. Finding the threshold between the two is essential. Therefore, our experiments measured sizeable real data sets to find the optimal values for this complex problem. The resulting algorithm can significantly facilitate the lengthy manual scheduling processes carried out so far in our university.
{"title":"Exploring the Potential of a Genetic Algorithm on a Real-World Complex Scheduling Problem","authors":"Szilvia Jáhn-Erdös, Bence Kövári","doi":"10.1109/ISCMI56532.2022.10068465","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068465","url":null,"abstract":"Genetic algorithms for NP-complete problems are widespread since it is easy to obtain a solution to the problem. However, its optimality gives the real issue and is not guaranteed to be achievable. In our research, we address a special subproblem of scheduling problems, the final exam scheduling, in which special requirements restrict the state space, which often contradicts each other. The task's difficulty is the massive size of the state space. Genetic algorithm-based solutions were considered since a MILP solver could not find a solution in a reasonable time. A model was built to solve this problem using the genetic algorithm. Most of the possibilities were seen in the different mutation procedures, so we investigated them in more detail. A question for genetic algorithms is what parameters and probabilities to run the model with since the more freedom we give to the run, the larger the runtime. Finding the threshold between the two is essential. Therefore, our experiments measured sizeable real data sets to find the optimal values for this complex problem. The resulting algorithm can significantly facilitate the lengthy manual scheduling processes carried out so far in our university.","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":"128649661","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/ISCMI56532.2022.10068442
Jerome Branny, Rolf Dornberger, T. Hanne
In this paper, we investigate how to forecast Non-Fungible Token (NFT) sale prices by using multiple multivariate time series datasets containing features related to the NFT market space. We examined eight recent studies regarding the forecasting and valuation of NFTs and compared their most important findings. This laid the fundamental work for two separate machine learning prototypes based on Long Short-Term Memory (LSTM) which are able to forecast the sale price history of an individual NFT asset. Root Mean Squared Errors (RMSE) of 0.2975 and 0.24 were obtained which appears to be promising.
{"title":"Non-fungible Token Price Prediction with Multivariate LSTM Neural Networks","authors":"Jerome Branny, Rolf Dornberger, T. Hanne","doi":"10.1109/ISCMI56532.2022.10068442","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068442","url":null,"abstract":"In this paper, we investigate how to forecast Non-Fungible Token (NFT) sale prices by using multiple multivariate time series datasets containing features related to the NFT market space. We examined eight recent studies regarding the forecasting and valuation of NFTs and compared their most important findings. This laid the fundamental work for two separate machine learning prototypes based on Long Short-Term Memory (LSTM) which are able to forecast the sale price history of an individual NFT asset. Root Mean Squared Errors (RMSE) of 0.2975 and 0.24 were obtained which appears to be promising.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"12 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":"128681183","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/ISCMI56532.2022.10068483
T. Dinh, Binh Pham Thanh
Lending activities are an important part of the credit activities of financial institutions and banks. This is an area that brings great potential for development as well as a sustainable source of profit for financial institutions and banks. However, lending to customers also brings high risks. Therefore, predicting the ability to repay on time and understanding the factors affecting the repayment ability of customers is extremely important and necessary, to help financial institutions and banks enhance their ability to pay debts. customers' ability to identify and pay debts on time, contributing to minimizing bad debts and enhancing credit risk management. In this study, Machine Learning models will be used: Proposing a method to combine Logistic Regression with Random Forest, Logistic Regression with K-Nearest Neighbor, Logistic Regression with Support Vector Machine, Logistic Regression with Artificial Neural Network, Logistic Regression with Long short-term memory and finally Logistic Regression with Decision Tree to predict customers' ability to repay on time and compare and evaluate the performance of Machine Learning models. As a result, the Logistic Regression with the Random Forest model ensemble is found as the optimal predictive model and it is expected that Fico Score and annual income significantly influence the forecast.
{"title":"Loan Repayment Prediction Using Logistic Regression Ensemble Learning With Machine Learning Algorithms","authors":"T. Dinh, Binh Pham Thanh","doi":"10.1109/ISCMI56532.2022.10068483","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068483","url":null,"abstract":"Lending activities are an important part of the credit activities of financial institutions and banks. This is an area that brings great potential for development as well as a sustainable source of profit for financial institutions and banks. However, lending to customers also brings high risks. Therefore, predicting the ability to repay on time and understanding the factors affecting the repayment ability of customers is extremely important and necessary, to help financial institutions and banks enhance their ability to pay debts. customers' ability to identify and pay debts on time, contributing to minimizing bad debts and enhancing credit risk management. In this study, Machine Learning models will be used: Proposing a method to combine Logistic Regression with Random Forest, Logistic Regression with K-Nearest Neighbor, Logistic Regression with Support Vector Machine, Logistic Regression with Artificial Neural Network, Logistic Regression with Long short-term memory and finally Logistic Regression with Decision Tree to predict customers' ability to repay on time and compare and evaluate the performance of Machine Learning models. As a result, the Logistic Regression with the Random Forest model ensemble is found as the optimal predictive model and it is expected that Fico Score and annual income significantly influence the forecast.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"13 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":"133938416","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/ISCMI56532.2022.10068459
Méziane Aïder, Imene Dey, M. Hifi
In this paper, we investigate the use of a population-based algorithm for tackling the fuzzy capacitated maximal covering location problem. Such a problem is characterized by a set of customers with their distances and its goal is to determine a subset of locations positioned on customers such that a maximum coverage of customers, including the both fuzzy coverage degree of facilities and the distance between customers, should be optimized. The proposed method is based upon the grey wolf optimizer, which starts by generating an initial population using a greedy rule strategy that is able to achieve feasible solutions according to the current positions of wolves. In order to enhance the quality of solutions induced, a series of local searches are added for exploring the search space by exploiting some nice strategies. The behavior of the method is computationally analyzed on a set of instances of the literature. Encouraging results have been provided.
{"title":"A Population-Based Algorithm for Solving the Fuzzy Capacitated Maximal Covering Location Problem","authors":"Méziane Aïder, Imene Dey, M. Hifi","doi":"10.1109/ISCMI56532.2022.10068459","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068459","url":null,"abstract":"In this paper, we investigate the use of a population-based algorithm for tackling the fuzzy capacitated maximal covering location problem. Such a problem is characterized by a set of customers with their distances and its goal is to determine a subset of locations positioned on customers such that a maximum coverage of customers, including the both fuzzy coverage degree of facilities and the distance between customers, should be optimized. The proposed method is based upon the grey wolf optimizer, which starts by generating an initial population using a greedy rule strategy that is able to achieve feasible solutions according to the current positions of wolves. In order to enhance the quality of solutions induced, a series of local searches are added for exploring the search space by exploiting some nice strategies. The behavior of the method is computationally analyzed on a set of instances of the literature. Encouraging results have been provided.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"171 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":"132182983","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/ISCMI56532.2022.10068484
Jasper Kyle Catapang
In this paper, the author proposes a new transformer model called Hadamard Estimated Attention Transformer or HEAT, that utilizes a low-rank projection of the Hadamard product to approximate the self-attention mechanism in standard transformer architectures and thus aiming to speedup transformer training, finetuning, and inference altogether. The study shows how it is significantly better than the original transformer that uses dot product self-attention by offering a faster way to compute the original self-attention mechanism while maintaining and ultimately surpassing the quality of the original transformer architecture. It also bests Linformer and Nyströmformer in several machine translation tasks while matching and even outperforming Nyströmformer's accuracy in various text classification tasks.
{"title":"Hadamard Estimated Attention Transformer (HEAT): Fast Approximation of Dot Product Self-attention for Transformers Using Low-Rank Projection of Hadamard Product","authors":"Jasper Kyle Catapang","doi":"10.1109/ISCMI56532.2022.10068484","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068484","url":null,"abstract":"In this paper, the author proposes a new transformer model called Hadamard Estimated Attention Transformer or HEAT, that utilizes a low-rank projection of the Hadamard product to approximate the self-attention mechanism in standard transformer architectures and thus aiming to speedup transformer training, finetuning, and inference altogether. The study shows how it is significantly better than the original transformer that uses dot product self-attention by offering a faster way to compute the original self-attention mechanism while maintaining and ultimately surpassing the quality of the original transformer architecture. It also bests Linformer and Nyströmformer in several machine translation tasks while matching and even outperforming Nyströmformer's accuracy in various text classification tasks.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"316 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":"134507700","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/ISCMI56532.2022.10068481
I. Rahimi, Theodore Picard, Andrew Morabito, Kiriakos Pampalis, Aiden Abignano, A. Gandomi
The problem of open-pit mining optimization is a complex task, often containing many variables. In this paper, we apply a trajectory-based algorithm known as simulated annealing together with a well-known population-based algorithm, genetic algorithm, used to generate solutions for a formulation of the constrained pit problem (CPIT). Three datasets were used to test this simulation, Newman1, zuck_small, and KD. The results show that simulated annealing as a trajectory algorithm possesses a slightly better performance in comparison with the genetic algorithm in terms of profit value.
{"title":"Comparison of Trajectory and Population-Based Algorithms for Optimizing Constrained Open-Pit Mining Problem","authors":"I. Rahimi, Theodore Picard, Andrew Morabito, Kiriakos Pampalis, Aiden Abignano, A. Gandomi","doi":"10.1109/ISCMI56532.2022.10068481","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068481","url":null,"abstract":"The problem of open-pit mining optimization is a complex task, often containing many variables. In this paper, we apply a trajectory-based algorithm known as simulated annealing together with a well-known population-based algorithm, genetic algorithm, used to generate solutions for a formulation of the constrained pit problem (CPIT). Three datasets were used to test this simulation, Newman1, zuck_small, and KD. The results show that simulated annealing as a trajectory algorithm possesses a slightly better performance in comparison with the genetic algorithm in terms of profit value.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"53 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":"124997707","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/ISCMI56532.2022.10068463
Abel S. Zacarias, L. Alexandre
Lifelong learning aims to develop machine learning systems that can learn new tasks while preserving the performance on previously learned tasks. Learning new tasks in most proposals, implies to keeping examples of previously learned tasks to retrain the model when learning new tasks, which has an impact in terms of storage capacity. In this paper, we present a method that adds new capabilities, in an incrementally way, to an existing model keeping examples from previously learned classes but avoiding the problem of running out of storage by using distilled images to condensate sets of images into a single image. The experimental results on four data sets confirmed the effectiveness of CILDI to learn new classes incrementally across different tasks and obtaining a performance close to the state-of-the-art algorithms for class incremental learning using only one distilled image per learned class and beating the state-of-the-art on the four data sets when using 10 distilled images per learned class, while using a smaller memory footprint than the competing approaches.
{"title":"CILDI: Class Incremental Learning with Distilled Images","authors":"Abel S. Zacarias, L. Alexandre","doi":"10.1109/ISCMI56532.2022.10068463","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068463","url":null,"abstract":"Lifelong learning aims to develop machine learning systems that can learn new tasks while preserving the performance on previously learned tasks. Learning new tasks in most proposals, implies to keeping examples of previously learned tasks to retrain the model when learning new tasks, which has an impact in terms of storage capacity. In this paper, we present a method that adds new capabilities, in an incrementally way, to an existing model keeping examples from previously learned classes but avoiding the problem of running out of storage by using distilled images to condensate sets of images into a single image. The experimental results on four data sets confirmed the effectiveness of CILDI to learn new classes incrementally across different tasks and obtaining a performance close to the state-of-the-art algorithms for class incremental learning using only one distilled image per learned class and beating the state-of-the-art on the four data sets when using 10 distilled images per learned class, while using a smaller memory footprint than the competing approaches.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"10 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":"125309795","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/ISCMI56532.2022.10068458
Ramona Boeh, T. Hanne, Rolf Dornberger
We compare the two parent selection methods “linear rank” and “tournament” in a Genetic Algorithm applied to a dynamic Travelling Salesman Problem (TSP). The inherent dynamics of the problem is considered by temporarily doubling the costs between two randomly selected cities. In our experiments we take into account tournament selection with tournament sizes of 3, 5, and 10. A larger tournament size results in as good a performance as with linear rank selection in a small-scale dynamic TSP, whereas smaller tournament sizes better preserve the diversity of the population and avoid getting stuck in local optima. However, the assumption that tournament is superior to linear rank on a dynamic TSP could neither be confirmed nor falsified in the applied testcases.
{"title":"A Comparison of Linear Rank and Tournament for Parent Selection in a Genetic Algorithm Solving a Dynamic Travelling Salesman Problem","authors":"Ramona Boeh, T. Hanne, Rolf Dornberger","doi":"10.1109/ISCMI56532.2022.10068458","DOIUrl":"https://doi.org/10.1109/ISCMI56532.2022.10068458","url":null,"abstract":"We compare the two parent selection methods “linear rank” and “tournament” in a Genetic Algorithm applied to a dynamic Travelling Salesman Problem (TSP). The inherent dynamics of the problem is considered by temporarily doubling the costs between two randomly selected cities. In our experiments we take into account tournament selection with tournament sizes of 3, 5, and 10. A larger tournament size results in as good a performance as with linear rank selection in a small-scale dynamic TSP, whereas smaller tournament sizes better preserve the diversity of the population and avoid getting stuck in local optima. However, the assumption that tournament is superior to linear rank on a dynamic TSP could neither be confirmed nor falsified in the applied testcases.","PeriodicalId":340397,"journal":{"name":"2022 9th International Conference on Soft Computing & Machine Intelligence (ISCMI)","volume":"5 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":"123899392","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：