Pub Date : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660094
Na'Shea Wiesner, John W. Sheppard, B. Haberman
This paper presents the results of experiments applying a Particle Swarm Optimization (PSO) approach to lane changing for autonomous vehicles. The lane change model proposed is rule-based, where PSO learns the parameters of the rules. A study was conducted to compare the proposed lane change model to the existing lane change model in the microscopic simulator, SUMO. Experiments performed include simulating vehicles using the Krauss car-following model with the SUMO lane change model, with the proposed PSO lane change model, and with all lane changing decisions turned off. The latter case, where merges are replaced by vehicle reset, serves as a baseline for missed merge opportunities. The objective was to develop an adaptive approach to improve merge efficiency as an example of lane changing behavior. Varying vehicle densities and levels of congestion on the merge lane and through-lane were tested. Empirical results show the proposed lane change model is able to learn merging strategies with minimal collisions and is comparable to the SUMO lane change model in some scenarios. Further investigation is needed to improve performance and safety, but initial results show promise for the proposed PSO-based approach to autonomous lane changing.
{"title":"Using Particle Swarm Optimization to Learn a Lane Change Model for Autonomous Vehicle Merging","authors":"Na'Shea Wiesner, John W. Sheppard, B. Haberman","doi":"10.1109/SSCI50451.2021.9660094","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660094","url":null,"abstract":"This paper presents the results of experiments applying a Particle Swarm Optimization (PSO) approach to lane changing for autonomous vehicles. The lane change model proposed is rule-based, where PSO learns the parameters of the rules. A study was conducted to compare the proposed lane change model to the existing lane change model in the microscopic simulator, SUMO. Experiments performed include simulating vehicles using the Krauss car-following model with the SUMO lane change model, with the proposed PSO lane change model, and with all lane changing decisions turned off. The latter case, where merges are replaced by vehicle reset, serves as a baseline for missed merge opportunities. The objective was to develop an adaptive approach to improve merge efficiency as an example of lane changing behavior. Varying vehicle densities and levels of congestion on the merge lane and through-lane were tested. Empirical results show the proposed lane change model is able to learn merging strategies with minimal collisions and is comparable to the SUMO lane change model in some scenarios. Further investigation is needed to improve performance and safety, but initial results show promise for the proposed PSO-based approach to autonomous lane changing.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"26 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":"129297405","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/SSCI50451.2021.9659943
Upasana Pattnaik, Minwoo Lee
Adapting feedback-driven deep reinforcement learning (DRL) algorithms to real-world problems requires developing robust systems that balance generalization and specialization. DRL algorithms powered by deep neural network function approximation tend to over-fit and perform poorly in new situations. Multi-task learning is a popular approach to reduce over-fitting by increasing input diversity, which in turn improves generalization capabilities. However, optimizing for multiple tasks often leads to distraction and performance oscillation. In this work, transfer learning paradigm Practice is introduced as an auxiliary task to stabilize distributed multi-task learning and enhance generalization. Experimental results demonstrate that the DRL algorithm supplemented with the state dynamics information produced by Practice improves performance.
{"title":"Multi-task Transfer with Practice","authors":"Upasana Pattnaik, Minwoo Lee","doi":"10.1109/SSCI50451.2021.9659943","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659943","url":null,"abstract":"Adapting feedback-driven deep reinforcement learning (DRL) algorithms to real-world problems requires developing robust systems that balance generalization and specialization. DRL algorithms powered by deep neural network function approximation tend to over-fit and perform poorly in new situations. Multi-task learning is a popular approach to reduce over-fitting by increasing input diversity, which in turn improves generalization capabilities. However, optimizing for multiple tasks often leads to distraction and performance oscillation. In this work, transfer learning paradigm Practice is introduced as an auxiliary task to stabilize distributed multi-task learning and enhance generalization. Experimental results demonstrate that the DRL algorithm supplemented with the state dynamics information produced by Practice improves performance.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"67 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":"130930225","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}
Although deep learning image segmentation technology has achieved good results in medical image processing, it is still challenging to segment renal parenchyma from diuretic renography. The diuretic nephrogram has the characteristics of obvious noise, poor image quality, unclear boundary and serious redundant information. It is difficult to accurately segment renal parenchyma directly using the classical Unet network. Therefore, we propose a cascaded network, i.e. a segment network that realize segmentation from coarse to fine. The coarse segmentation model is used to obtain the suggested area of the kidney in the diuretic renal image. The cascaded fine segmentation model is to segment the renal parenchyma from the suggested region of the kidney. Compared with the original Unet, the cascade network can reduce the noise interference to a large extent and get better segmentation performance of the renal parenchyma. The experiment showed that the dice coefficient increased by 9.78%, and the proposed network is efficient in the renal parenchyma segmentation.
{"title":"UUnet: An effective cascade Unet for automatic segmentation of renal parenchyma","authors":"Gaoyu Cao, Zhanquan Sun, Minlan Pan, Jiangfei Pang, Zhiqiang He, Jiayu Shen","doi":"10.1109/SSCI50451.2021.9660077","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660077","url":null,"abstract":"Although deep learning image segmentation technology has achieved good results in medical image processing, it is still challenging to segment renal parenchyma from diuretic renography. The diuretic nephrogram has the characteristics of obvious noise, poor image quality, unclear boundary and serious redundant information. It is difficult to accurately segment renal parenchyma directly using the classical Unet network. Therefore, we propose a cascaded network, i.e. a segment network that realize segmentation from coarse to fine. The coarse segmentation model is used to obtain the suggested area of the kidney in the diuretic renal image. The cascaded fine segmentation model is to segment the renal parenchyma from the suggested region of the kidney. Compared with the original Unet, the cascade network can reduce the noise interference to a large extent and get better segmentation performance of the renal parenchyma. The experiment showed that the dice coefficient increased by 9.78%, and the proposed network is efficient in the renal parenchyma segmentation.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"106 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":"131713705","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/SSCI50451.2021.9660084
Kyle W. Pretorius, N. Pillay
Genetic algorithms have recently seen an increase in application due to their highly scalable nature. Enabling more efficient utilization of processing power that has become readily available. This study introduces Population based reinforcement learning (PBRL), a method that hybridizes a GA with a policy gradient reinforcement learning algorithm. This combination not only enables more scalable policy optimization, but also helps mitigate some of the common weaknesses of policy gradient algorithms. Furthermore, PBRL is also extended to include automatic hyper-parameter tuning, which is used to evaluate the impact that such tuning can have on the performance of the policy gradient algorithm being used. Experiments comparing these methods are conducted on a number of continuous control problems simulated by MuJoCo. Results show that PBRL is capable of outperforming a commonly used policy gradient algorithm, while also producing results in nearly one fifth the time. It is also observed that the addition of automatic hyperparameter tuning can be greatly beneficial for environments where well tuned hyper-parameters are not known.
{"title":"Population based Reinforcement Learning","authors":"Kyle W. Pretorius, N. Pillay","doi":"10.1109/SSCI50451.2021.9660084","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660084","url":null,"abstract":"Genetic algorithms have recently seen an increase in application due to their highly scalable nature. Enabling more efficient utilization of processing power that has become readily available. This study introduces Population based reinforcement learning (PBRL), a method that hybridizes a GA with a policy gradient reinforcement learning algorithm. This combination not only enables more scalable policy optimization, but also helps mitigate some of the common weaknesses of policy gradient algorithms. Furthermore, PBRL is also extended to include automatic hyper-parameter tuning, which is used to evaluate the impact that such tuning can have on the performance of the policy gradient algorithm being used. Experiments comparing these methods are conducted on a number of continuous control problems simulated by MuJoCo. Results show that PBRL is capable of outperforming a commonly used policy gradient algorithm, while also producing results in nearly one fifth the time. It is also observed that the addition of automatic hyperparameter tuning can be greatly beneficial for environments where well tuned hyper-parameters are not known.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"17 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":"126638960","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/SSCI50451.2021.9659882
Stefano Iannucci, E. Casalicchio, Matteo Lucantonio
Intrusion Response is a relatively new field of research. Several model-based techniques have been proposed that range from static mapping to complex stateful approaches. However, the main limitation that all of them have in common is that they do not consider the non-stationary behavior of the protected system which, in combination with long planning times, makes it unfeasible to use them on dynamic and large-scale systems. In this work, we propose an Intrusion Response controller based on deep reinforcement learning and transfer learning, which automatically adapts to system changes. We empirically demonstrate its effectiveness and its performance on Online Boutique, a cloud-based web application that Google uses to showcase its cloud technologies. We first carry out an extensive tuning of the hyper-parameters of the neural networks that implement our approach. Afterwards, we empirically show the effectiveness and the performance of the realized Intrusion Response controller in a typical cloud scenario, that is, when instances are added or removed from the system. Experimental results show that a proper hyper-parameter tuning can reduce the training time by up to 50%. Furthermore, transfer learning completely zeroes the transient adaptation stage when the number of replicas of a given service is reduced. The training during the transient stage exhibits instead a speed-up of 1.25x in case a replica is added. For reproducibility, the source code of the Intrusion Response System is released with the onen-source Apache 2.0 license.
{"title":"An Intrusion Response Approach for Elastic Applications Based on Reinforcement Learning","authors":"Stefano Iannucci, E. Casalicchio, Matteo Lucantonio","doi":"10.1109/SSCI50451.2021.9659882","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659882","url":null,"abstract":"Intrusion Response is a relatively new field of research. Several model-based techniques have been proposed that range from static mapping to complex stateful approaches. However, the main limitation that all of them have in common is that they do not consider the non-stationary behavior of the protected system which, in combination with long planning times, makes it unfeasible to use them on dynamic and large-scale systems. In this work, we propose an Intrusion Response controller based on deep reinforcement learning and transfer learning, which automatically adapts to system changes. We empirically demonstrate its effectiveness and its performance on Online Boutique, a cloud-based web application that Google uses to showcase its cloud technologies. We first carry out an extensive tuning of the hyper-parameters of the neural networks that implement our approach. Afterwards, we empirically show the effectiveness and the performance of the realized Intrusion Response controller in a typical cloud scenario, that is, when instances are added or removed from the system. Experimental results show that a proper hyper-parameter tuning can reduce the training time by up to 50%. Furthermore, transfer learning completely zeroes the transient adaptation stage when the number of replicas of a given service is reduced. The training during the transient stage exhibits instead a speed-up of 1.25x in case a replica is added. For reproducibility, the source code of the Intrusion Response System is released with the onen-source Apache 2.0 license.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"72 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":"126445910","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/SSCI50451.2021.9659997
A. Zaiou, Younès Bennani, Basarab Matei, M. Hibti
In this paper, we propose a new quantum version of the Balanced K-means algorithm in the D-wave quantum annealer. D-wave 2000Q quantum computer has been used by many papers in the last few years to solve optimization problems and for finding the global minimum of the balanced K-means optimization problem. However, in this paper, we modify the quadratic unconstrained binary optimization (QUBO) formulation of the Balanced K-means that has been proposed in a recent paper. Our modification is trained on different data sets: Iris, Wine and Breast Cancer. Also, we performed a comparative analysis between the two approaches (our approach and the paper's approach) to find the one that assigns the largest number of data to clusters and we also use the Davies-Bouldi metric to prove that our method gives the best clustering.
{"title":"Balanced K-means using Quantum annealing","authors":"A. Zaiou, Younès Bennani, Basarab Matei, M. Hibti","doi":"10.1109/SSCI50451.2021.9659997","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659997","url":null,"abstract":"In this paper, we propose a new quantum version of the Balanced K-means algorithm in the D-wave quantum annealer. D-wave 2000Q quantum computer has been used by many papers in the last few years to solve optimization problems and for finding the global minimum of the balanced K-means optimization problem. However, in this paper, we modify the quadratic unconstrained binary optimization (QUBO) formulation of the Balanced K-means that has been proposed in a recent paper. Our modification is trained on different data sets: Iris, Wine and Breast Cancer. Also, we performed a comparative analysis between the two approaches (our approach and the paper's approach) to find the one that assigns the largest number of data to clusters and we also use the Davies-Bouldi metric to prove that our method gives the best clustering.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"18 5","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114131599","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/SSCI50451.2021.9659927
Gbenga Omotara, Mark L. Berardi, Maria Dietrich, G. DeSouza
Relative fundamental frequency (RFF) is an acoustic measure used to quantify vocal effort in voice science. Since it seeks to capture transitions between (i.e. to/from) steady-state vowels and unvoiced consonants, any machine learning approach to recognize patterns in these transitions should require time properties capable of identifying the sequence of phonemes. At the same time, Neural Networks (NN) have become a ubiquitous solution for data-driven problems, and Recursive NNs (RNN) provide a time-series schema to address time-dependent problems. Indeed, typical Neural Network solutions require either a time-series schema like in RNN or some spectral transformation to be able to handle time-dependent data. In this study, we decided to ignore - at least momentarily - any time-series dependency of the data and employed a simple NN to classify elements of the speech. Later, a State-Machine was used to identify their sequence with the purpose of localizing the transitions between voiced and unvoiced sounds in vowel-consonant-vowel (VCV) productions. The goal of this study was to demonstrate that a pipeline consisting of time-agnostic (Neural Network) and time-dependent (State Machine) components can be used to recognize time-dependent patterns in VCV productions.
{"title":"A Pipeline Consisting of Pattern Recognition and Finite Automata for Recognizing VCV Productions in the Study of Vocal Hyperfunction","authors":"Gbenga Omotara, Mark L. Berardi, Maria Dietrich, G. DeSouza","doi":"10.1109/SSCI50451.2021.9659927","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659927","url":null,"abstract":"Relative fundamental frequency (RFF) is an acoustic measure used to quantify vocal effort in voice science. Since it seeks to capture transitions between (i.e. to/from) steady-state vowels and unvoiced consonants, any machine learning approach to recognize patterns in these transitions should require time properties capable of identifying the sequence of phonemes. At the same time, Neural Networks (NN) have become a ubiquitous solution for data-driven problems, and Recursive NNs (RNN) provide a time-series schema to address time-dependent problems. Indeed, typical Neural Network solutions require either a time-series schema like in RNN or some spectral transformation to be able to handle time-dependent data. In this study, we decided to ignore - at least momentarily - any time-series dependency of the data and employed a simple NN to classify elements of the speech. Later, a State-Machine was used to identify their sequence with the purpose of localizing the transitions between voiced and unvoiced sounds in vowel-consonant-vowel (VCV) productions. The goal of this study was to demonstrate that a pipeline consisting of time-agnostic (Neural Network) and time-dependent (State Machine) components can be used to recognize time-dependent patterns in VCV productions.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"68 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":"114266855","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/SSCI50451.2021.9660087
Lin Teng, Nobuhiro Watanabe, T. Izumi, H. Matsuzaki, Tomoaki Takahashi, N. Kiryu
In traffic control systems, the vehicle detectors play a central role in collecting information. As information gathering equipment, there are many types of current vehicle detectors, such as ultrasonic type, loop coil type, and image type. But there are problems such as reduced accuracy due to temperature and wind, maintainability problems, and reduced accuracy at night. Therefore, in this research, we are considering using a Laser Ranging Image Sensor that can obtain an image including the distance as a new vehicle detector. In this paper, we first examined the sensor installation conditions. And then we proposed vehicle detection method for Laser Ranging Image Sensor for One-lane detection and Multiple-lanes detection. Finally, we conducted the demonstration experiment of vehicle detection and compared with the few public information of traditional vehicle detectors. From experiment results, we confirmed that for One-lane detection, it is excellent in accuracy. For Multiple-lines detection, it also has sufficient detection accuracy and can be further improving. In addition, as a new vehicle detector, it is excellent in installation and maintenance than traditional vehicle detectors.
{"title":"Application on Vehicle Detector using Laser Ranging Image Sensor","authors":"Lin Teng, Nobuhiro Watanabe, T. Izumi, H. Matsuzaki, Tomoaki Takahashi, N. Kiryu","doi":"10.1109/SSCI50451.2021.9660087","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660087","url":null,"abstract":"In traffic control systems, the vehicle detectors play a central role in collecting information. As information gathering equipment, there are many types of current vehicle detectors, such as ultrasonic type, loop coil type, and image type. But there are problems such as reduced accuracy due to temperature and wind, maintainability problems, and reduced accuracy at night. Therefore, in this research, we are considering using a Laser Ranging Image Sensor that can obtain an image including the distance as a new vehicle detector. In this paper, we first examined the sensor installation conditions. And then we proposed vehicle detection method for Laser Ranging Image Sensor for One-lane detection and Multiple-lanes detection. Finally, we conducted the demonstration experiment of vehicle detection and compared with the few public information of traditional vehicle detectors. From experiment results, we confirmed that for One-lane detection, it is excellent in accuracy. For Multiple-lines detection, it also has sufficient detection accuracy and can be further improving. In addition, as a new vehicle detector, it is excellent in installation and maintenance than traditional vehicle detectors.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"19 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":"125138740","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/SSCI50451.2021.9660031
Fabian Hinder, Valerie Vaquet, Johannes Brinkrolf, Barbara Hammer
In many machine learning tasks, one tries to infer unknown quantities such as the conditional density p(Y | X) from observed ones X. Conditional density estimation (CDE) constitutes a challenging problem due to the trade-off between model complexity, distribution complexity, and overfitting. In case of online learning, where the distribution may change over time (concept drift) or only few data points are available at once, robust, non-parametric approaches are of particular interest. In this paper we present a new, non-parametric tree-ensemble-based method for CDE that reduces the problem to a simple regression task on the transformed input data and a (unconditional) density estimation. We prove the correctness of our approach and show its usefulness in empirical evaluation on standard benchmarks. We show that our method is comparable to other state-of-the-art methods, but is much faster and more robust.
{"title":"Fast Non-Parametric Conditional Density Estimation using Moment Trees","authors":"Fabian Hinder, Valerie Vaquet, Johannes Brinkrolf, Barbara Hammer","doi":"10.1109/SSCI50451.2021.9660031","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660031","url":null,"abstract":"In many machine learning tasks, one tries to infer unknown quantities such as the conditional density p(Y | X) from observed ones X. Conditional density estimation (CDE) constitutes a challenging problem due to the trade-off between model complexity, distribution complexity, and overfitting. In case of online learning, where the distribution may change over time (concept drift) or only few data points are available at once, robust, non-parametric approaches are of particular interest. In this paper we present a new, non-parametric tree-ensemble-based method for CDE that reduces the problem to a simple regression task on the transformed input data and a (unconditional) density estimation. We prove the correctness of our approach and show its usefulness in empirical evaluation on standard benchmarks. We show that our method is comparable to other state-of-the-art methods, but is much faster and more robust.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"102 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":"131617299","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/SSCI50451.2021.9660053
Nicholas Masri, M. Yetkin, Emma Hillman, Daniel Fay, S. Kishore
This paper presents a mathematical model to optimize the depot charging of a transit electric bus fleet given route schedules. The total cost of electricity drawn from the power grid is minimized in the presence of solar (PV) generation and an energy storage system (ESS). A mixed-integer linear program (MILP) is proposed that allows for static bus-route pairs as well as more flexible and dynamic assignments of buses to routes. The models are tested on data from a transit agency in Santa Clara County, California, including time-of-use (TOU) grid prices, vehicle specifications, and solar generation capabilities. Results show both models achieve full solar utilization and consume less power during peak hours. Dynamic route assignment achieves an 11% reduction in operational costs and performs over a range of parameters based on sensitivity analysis. Furthermore, the results demonstrate the effect of weather on operational costs and other operational strategies.
{"title":"Optimal Public Electric Bus Fleet Charging Schedule with Solar and Energy Storage Considering Static and Dynamic Route Assignment","authors":"Nicholas Masri, M. Yetkin, Emma Hillman, Daniel Fay, S. Kishore","doi":"10.1109/SSCI50451.2021.9660053","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660053","url":null,"abstract":"This paper presents a mathematical model to optimize the depot charging of a transit electric bus fleet given route schedules. The total cost of electricity drawn from the power grid is minimized in the presence of solar (PV) generation and an energy storage system (ESS). A mixed-integer linear program (MILP) is proposed that allows for static bus-route pairs as well as more flexible and dynamic assignments of buses to routes. The models are tested on data from a transit agency in Santa Clara County, California, including time-of-use (TOU) grid prices, vehicle specifications, and solar generation capabilities. Results show both models achieve full solar utilization and consume less power during peak hours. Dynamic route assignment achieves an 11% reduction in operational costs and performs over a range of parameters based on sensitivity analysis. Furthermore, the results demonstrate the effect of weather on operational costs and other operational strategies.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"26 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":"127817063","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：