Pub Date : 2021-12-05DOI: 10.1109/SSCI50451.2021.9660126
Gang Li, Ogechi Onuoha, Mark Mcgill, S. Brewster, C. Chen, F. Pollick
How the performance of autonomic physiological, and human vestibular network (HVN)-based brain functional connectivity (BFC) features differ in a virtual reality (VR) sickness classification task is underexplored. Therefore, this paper presents an artificial intelligence (AI)-aided comparative study of the two. Results from different AI models all show that autonomic physiological features represented by the combined heart rate, fingertip temperature and forehead temperature are superior to HVN-based BFC features represented by the phase-locking values of inter-electrode coherence (IEC) of electroencephalogram (EEG) in the same VR sickness condition (that is, as a result of experiencing tunnel travel-induced illusory self-motion (vection) about moving in-depth in this study). Regarding EEG features per se (IEC-BFC vs traditional power spectrum), we did not find much difference across AI models.
{"title":"Comparing Autonomic Physiological and Electroencephalography Features for VR Sickness Detection Using Predictive Models","authors":"Gang Li, Ogechi Onuoha, Mark Mcgill, S. Brewster, C. Chen, F. Pollick","doi":"10.1109/SSCI50451.2021.9660126","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660126","url":null,"abstract":"How the performance of autonomic physiological, and human vestibular network (HVN)-based brain functional connectivity (BFC) features differ in a virtual reality (VR) sickness classification task is underexplored. Therefore, this paper presents an artificial intelligence (AI)-aided comparative study of the two. Results from different AI models all show that autonomic physiological features represented by the combined heart rate, fingertip temperature and forehead temperature are superior to HVN-based BFC features represented by the phase-locking values of inter-electrode coherence (IEC) of electroencephalogram (EEG) in the same VR sickness condition (that is, as a result of experiencing tunnel travel-induced illusory self-motion (vection) about moving in-depth in this study). Regarding EEG features per se (IEC-BFC vs traditional power spectrum), we did not find much difference across AI models.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"7 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":"126766508","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/SSCI50451.2021.9659838
Somayeh Bakhtiari Ramezani, Brad Killen, Logan Cummins, S. Rahimi, A. Amirlatifi, Maria Seale
Early detection of faulty patterns and timely scheduling of maintenance events can minimize risk to the underlying processes and increase the system's lifespan, reliability, and availability. Different techniques are used in the literature to determine the health state of the system, one of which is the Hidden Markov Models (HMMs). This class of algorithms is very well suited for modeling the health condition dictated by the latent states of the system. HMMs can reveal transitions from one state to another, thus highlighting degradation in a system's health and the right time for maintenance. While many extensions and variations of the HMM are studied for a variety of applications, the present study aims to evaluate and compare the state-of-the-art HMM-based research in predictive maintenance only. This study also aims to discuss the capabilities and limitations of such algorithms and future directions to tackle the current limitations.
{"title":"A Survey of HMM-based Algorithms in Machinery Fault Prediction","authors":"Somayeh Bakhtiari Ramezani, Brad Killen, Logan Cummins, S. Rahimi, A. Amirlatifi, Maria Seale","doi":"10.1109/SSCI50451.2021.9659838","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659838","url":null,"abstract":"Early detection of faulty patterns and timely scheduling of maintenance events can minimize risk to the underlying processes and increase the system's lifespan, reliability, and availability. Different techniques are used in the literature to determine the health state of the system, one of which is the Hidden Markov Models (HMMs). This class of algorithms is very well suited for modeling the health condition dictated by the latent states of the system. HMMs can reveal transitions from one state to another, thus highlighting degradation in a system's health and the right time for maintenance. While many extensions and variations of the HMM are studied for a variety of applications, the present study aims to evaluate and compare the state-of-the-art HMM-based research in predictive maintenance only. This study also aims to discuss the capabilities and limitations of such algorithms and future directions to tackle the current limitations.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"33 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":"127999600","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/SSCI50451.2021.9660081
Ganesh Tata, Eric Austin
The overestimation of action values caused by randomness in rewards can harm the ability to learn and the performance of reinforcement learning agents. This maximization bias has been well established and studied in the off-policy Q-learning algorithm. However, less study has been done for on-policy algorithms such as Sarsa and its variants. We conduct a thorough empirical analysis on Sarsa, Expected Sarsa, and n-step Sarsa. We find that the on-policy Sarsa variants suffer from less maximization bias than off-policy Q-learning in several test environments. We show how the choice of hyper-parameters impacts the severity of the bias. A decaying learning rate schedule results in more maximization bias than a fixed learning rate. Larger learning rates lead to larger overestimation. A larger exploration parameter leads to worse bias in Q-learning but less bias in the on-policy algorithms. We also show that a larger variance in rewards leads to more bias in both Q-Learning and Sarsa., but Sarsa is less affected than Q-learning.
{"title":"Investigation of Maximization Bias in Sarsa Variants","authors":"Ganesh Tata, Eric Austin","doi":"10.1109/SSCI50451.2021.9660081","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9660081","url":null,"abstract":"The overestimation of action values caused by randomness in rewards can harm the ability to learn and the performance of reinforcement learning agents. This maximization bias has been well established and studied in the off-policy Q-learning algorithm. However, less study has been done for on-policy algorithms such as Sarsa and its variants. We conduct a thorough empirical analysis on Sarsa, Expected Sarsa, and n-step Sarsa. We find that the on-policy Sarsa variants suffer from less maximization bias than off-policy Q-learning in several test environments. We show how the choice of hyper-parameters impacts the severity of the bias. A decaying learning rate schedule results in more maximization bias than a fixed learning rate. Larger learning rates lead to larger overestimation. A larger exploration parameter leads to worse bias in Q-learning but less bias in the on-policy algorithms. We also show that a larger variance in rewards leads to more bias in both Q-Learning and Sarsa., but Sarsa is less affected than Q-learning.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"85 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":"127654471","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/SSCI50451.2021.9659850
M. Alghamdi, P. Angelov, Bryan M. Williams
Handimages are of paramount importance within critical domains like security and criminal investigation. They can sometimes be the only available evidence of an offender's identity at a crime scene. Approaches to person identification that consider the human hand as a complex object composed of many components are rare. The approach proposed in this paper fills this gap, making use of knuckle creases and fingernail information. It introduces a framework for automatic person identification that includes localisation of the regions of interest within hand images, recognition of the detected components, segmentation of the region of interest using bounding boxes, and similarity matching between a query image and a library of available images. The following hand components are considered: i) the metacarpohalangeal, commonly known as base knuckle; ii) the proximal interphalangeal joint commonly known as major knuckle; iii) distal interphalangeal joint, commonly known as minor knuckle; iv) the interphalangeal joint, commonly known as thumb's knuckle, and v) the fingernails. A key element of the proposed framework is the similarity matching and an important role for it is played by the feature extraction. In this paper, we exploit end-to-end deep convolutional neural networks to extract discriminative high-level abstract features. We further use Bray-Curtis (BC) similarity for the matching process. We validated the proposed approach on well-known benchmarks, the ‘11k Hands' dataset and the Hong Kong Polytechnic University Contactless Hand Dorsal Images known as ‘PolyU HD’. We found that the results indicate that the knuckle patterns and fingernails play a significant role in the person identification. The results from the 11K dataset indicate that the results for the left hand are better than the results for the right hand. In both datasets, the fingernails produced consistently higher identification results than other hand components, with a rank-1 score of 93.65% on the ring finger of the left hand for the ‘11k Hands' dataset and rank-l score of 93.81% for the thumb from the ‘PolyU HD’ dataset.
{"title":"Automated Person Identification Framework Based on Fingernails and Dorsal Knuckle Patterns","authors":"M. Alghamdi, P. Angelov, Bryan M. Williams","doi":"10.1109/SSCI50451.2021.9659850","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659850","url":null,"abstract":"Handimages are of paramount importance within critical domains like security and criminal investigation. They can sometimes be the only available evidence of an offender's identity at a crime scene. Approaches to person identification that consider the human hand as a complex object composed of many components are rare. The approach proposed in this paper fills this gap, making use of knuckle creases and fingernail information. It introduces a framework for automatic person identification that includes localisation of the regions of interest within hand images, recognition of the detected components, segmentation of the region of interest using bounding boxes, and similarity matching between a query image and a library of available images. The following hand components are considered: i) the metacarpohalangeal, commonly known as base knuckle; ii) the proximal interphalangeal joint commonly known as major knuckle; iii) distal interphalangeal joint, commonly known as minor knuckle; iv) the interphalangeal joint, commonly known as thumb's knuckle, and v) the fingernails. A key element of the proposed framework is the similarity matching and an important role for it is played by the feature extraction. In this paper, we exploit end-to-end deep convolutional neural networks to extract discriminative high-level abstract features. We further use Bray-Curtis (BC) similarity for the matching process. We validated the proposed approach on well-known benchmarks, the ‘11k Hands' dataset and the Hong Kong Polytechnic University Contactless Hand Dorsal Images known as ‘PolyU HD’. We found that the results indicate that the knuckle patterns and fingernails play a significant role in the person identification. The results from the 11K dataset indicate that the results for the left hand are better than the results for the right hand. In both datasets, the fingernails produced consistently higher identification results than other hand components, with a rank-1 score of 93.65% on the ring finger of the left hand for the ‘11k Hands' dataset and rank-l score of 93.81% for the thumb from the ‘PolyU HD’ dataset.","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":"129200345","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/SSCI50451.2021.9659976
L. Glass, Wael Hilali, O. Nelles
We present Rectified Linear Unit based Local Linear Model Tree (ReLUMoT). A model that bridges the gap between Piecewise Linear Neural Networks (PLNN) and Local Model Networks (LMN) like those resulting from the LoLiMoT algorithm. Essentially, we perform the input space partitioning of LoLiMoT by training a PLNN and extracting its linear regions. These become the input space partitions of ReLUMoT. From the perspective of PLNNs our approach compresses and smoothens low-dimensional models, while making them interpretable. From the perspective of LoLiMoT, our approach replaces the incremental and heuristic input space partitioning scheme with gradient-based training of a neural network, which is considerably more flexible.
{"title":"Compressing Interpretable Representations of Piecewise Linear Neural Networks using Neuro-Fuzzy Models","authors":"L. Glass, Wael Hilali, O. Nelles","doi":"10.1109/SSCI50451.2021.9659976","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659976","url":null,"abstract":"We present Rectified Linear Unit based Local Linear Model Tree (ReLUMoT). A model that bridges the gap between Piecewise Linear Neural Networks (PLNN) and Local Model Networks (LMN) like those resulting from the LoLiMoT algorithm. Essentially, we perform the input space partitioning of LoLiMoT by training a PLNN and extracting its linear regions. These become the input space partitions of ReLUMoT. From the perspective of PLNNs our approach compresses and smoothens low-dimensional models, while making them interpretable. From the perspective of LoLiMoT, our approach replaces the incremental and heuristic input space partitioning scheme with gradient-based training of a neural network, which is considerably more flexible.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"4 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":"132653522","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/SSCI50451.2021.9659911
Carlos Quintero Gull, J. Aguilar, M. Rodríguez-Moreno
In this work, we use the semi-supervised LAMDA-HSCC algorithm for characterizing the energy consumption in smart buildings, which can work with labeled and unlabeled data. Particularly, it uses the LAMDA-RD approach for the clustering problem and the LAMDA-HAD approach for the classification problem. Additionally, this algorithm uses three submodels for merging, partition groups (classes/cluster) and migrating individuals from a group to another. For the performance evaluation, several datasets of energetic consumption are used, with different percent of labeled data, showing very encouraging results according to two metrics in the semi-supervised context.
{"title":"A semi-supervised learning approach to study the energy consumption in smart buildings","authors":"Carlos Quintero Gull, J. Aguilar, M. Rodríguez-Moreno","doi":"10.1109/SSCI50451.2021.9659911","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659911","url":null,"abstract":"In this work, we use the semi-supervised LAMDA-HSCC algorithm for characterizing the energy consumption in smart buildings, which can work with labeled and unlabeled data. Particularly, it uses the LAMDA-RD approach for the clustering problem and the LAMDA-HAD approach for the classification problem. Additionally, this algorithm uses three submodels for merging, partition groups (classes/cluster) and migrating individuals from a group to another. For the performance evaluation, several datasets of energetic consumption are used, with different percent of labeled data, showing very encouraging results according to two metrics in the semi-supervised context.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"51 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":"132298127","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/SSCI50451.2021.9659932
Campbell Gorman, Yu-kai Wang
Both Augmented Reality (AR) and Brain-Computer Interfaces (BCI) have drawn a lot of attention in recent applications. These two new technologies will significantly impact and develop interactions between human and intelligent agents. While there are several studies already conducted in the control of devices using AR based, steady state visually evoked potentials (SSVEP) control systems in a lab environment, this study seeks to implement a portable, closed-loop, AR-based BCI to assess the feasibility of controlling a physical device through SSVEP. This portable, closed-loop AR-based BCI provides users with the unique opportunity to simultaneously interact with the surrounding environment and control autonomous agents with an 88% accuracy. The potential benefits of this application include reduced restrictions on handicapped individuals or concurrent control of multiple devices through a single AR interface. Ultimately, we hope this outcome can bridge the BCI field with further real-world, practical applications.
{"title":"A Closed-Loop AR-based BCI for Real-World System Control","authors":"Campbell Gorman, Yu-kai Wang","doi":"10.1109/SSCI50451.2021.9659932","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659932","url":null,"abstract":"Both Augmented Reality (AR) and Brain-Computer Interfaces (BCI) have drawn a lot of attention in recent applications. These two new technologies will significantly impact and develop interactions between human and intelligent agents. While there are several studies already conducted in the control of devices using AR based, steady state visually evoked potentials (SSVEP) control systems in a lab environment, this study seeks to implement a portable, closed-loop, AR-based BCI to assess the feasibility of controlling a physical device through SSVEP. This portable, closed-loop AR-based BCI provides users with the unique opportunity to simultaneously interact with the surrounding environment and control autonomous agents with an 88% accuracy. The potential benefits of this application include reduced restrictions on handicapped individuals or concurrent control of multiple devices through a single AR interface. Ultimately, we hope this outcome can bridge the BCI field with further real-world, practical applications.","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":"130349926","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/SSCI50451.2021.9659928
Shirsha Bose, Sayantani Ghosh, A. Konar, A. Nagar
The paper introduces an innovative methodology for the automatic discrimination of multiple choice answers chosen by merit and random guess by analyzing the confidence level of examinees using an Electroencephalographic system. The acquired brain signals of the subjects participating in the experiment are first examined using the eLORETA software which portrays the active participation of the middle frontal gyrus and precuneus when a subject is fully confident regarding the choice of the correct answer. In the next phase, the signals are pre-processed and converted to spectrogram plots using Short Time Fourier Transform (STFT) which reveal the enhanced activation of theta and lower alpha bands when a subject attempts an answer with his/her merit. On the other hand, the afore-said frequency bands portray reduced activation when a subject tries to choose an answer by a mere guess. The acquired spectrogram plots are transferred to a novel Capsule network model that aids in categorizing the two degrees of confidence level: High and Low. The novelty in the design of the Capsule based classifier lies in the introduction of a depthwise separable convolution layer, a squeeze and excitation attention mechanism and a Sigmoid-Weighted Linear Unit (SiLU) based dynamic routing algorithm. The proposed classifier demonstrates promising results in categorizing the two classes of confidence level and also outperforms its conventional counterparts. Thus, the proposed scheme can be utilized to improve the quality of assessment in multiple choice based examinations.
{"title":"Decoding the Confidence Level of Subjects in Answering Multiple Choice Questions Using EEG Induced Capsule Network","authors":"Shirsha Bose, Sayantani Ghosh, A. Konar, A. Nagar","doi":"10.1109/SSCI50451.2021.9659928","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659928","url":null,"abstract":"The paper introduces an innovative methodology for the automatic discrimination of multiple choice answers chosen by merit and random guess by analyzing the confidence level of examinees using an Electroencephalographic system. The acquired brain signals of the subjects participating in the experiment are first examined using the eLORETA software which portrays the active participation of the middle frontal gyrus and precuneus when a subject is fully confident regarding the choice of the correct answer. In the next phase, the signals are pre-processed and converted to spectrogram plots using Short Time Fourier Transform (STFT) which reveal the enhanced activation of theta and lower alpha bands when a subject attempts an answer with his/her merit. On the other hand, the afore-said frequency bands portray reduced activation when a subject tries to choose an answer by a mere guess. The acquired spectrogram plots are transferred to a novel Capsule network model that aids in categorizing the two degrees of confidence level: High and Low. The novelty in the design of the Capsule based classifier lies in the introduction of a depthwise separable convolution layer, a squeeze and excitation attention mechanism and a Sigmoid-Weighted Linear Unit (SiLU) based dynamic routing algorithm. The proposed classifier demonstrates promising results in categorizing the two classes of confidence level and also outperforms its conventional counterparts. Thus, the proposed scheme can be utilized to improve the quality of assessment in multiple choice based examinations.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"6 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":"130421846","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/SSCI50451.2021.9659938
L. A. D. Silveira, J. L. Soncco-Álvarez, T. Lima, M. Ayala-Rincón
Homogeneous Parallel Island Models (HoPIMs) run the same bio-inspired algorithm (BA) in all islands. Several communication topologies and migration policies have been fine-tuned in such models, speeding up and providing better quality solutions than sequential BAs for different case studies. This work selects four HoPIMs that successfully ran a genetic algorithm (GA) in all their islands. Furthermore, it proposes and studies the performance of heterogeneous versions of such models (HePIMs) that run four different BAs in their islands, namely, GA, double-point crossover GA, Differential Evolution, and Particle Swarm Optimization. HePIMs aim to maintain population diversity covering the space of solutions and reducing the overlap between islands. The NP-hard evolutionary reversal distance problem is addressed with HePIMs verifying their ability to compute accurate solutions and outperforming HoPIMs.
{"title":"Heterogeneous Parallel Island Models","authors":"L. A. D. Silveira, J. L. Soncco-Álvarez, T. Lima, M. Ayala-Rincón","doi":"10.1109/SSCI50451.2021.9659938","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659938","url":null,"abstract":"Homogeneous Parallel Island Models (HoPIMs) run the same bio-inspired algorithm (BA) in all islands. Several communication topologies and migration policies have been fine-tuned in such models, speeding up and providing better quality solutions than sequential BAs for different case studies. This work selects four HoPIMs that successfully ran a genetic algorithm (GA) in all their islands. Furthermore, it proposes and studies the performance of heterogeneous versions of such models (HePIMs) that run four different BAs in their islands, namely, GA, double-point crossover GA, Differential Evolution, and Particle Swarm Optimization. HePIMs aim to maintain population diversity covering the space of solutions and reducing the overlap between islands. The NP-hard evolutionary reversal distance problem is addressed with HePIMs verifying their ability to compute accurate solutions and outperforming HoPIMs.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"254 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":"132429872","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/SSCI50451.2021.9659973
Pubudu Sanjeewani, B. Verma, J. Affum
The accurate recognition of road markings such as lanes and turn arrows is required in many applications including autonomous vehicles. Nevertheless, studies on road markings detection are commonly found in literature, detection and classification of turn lane arrows has not gained much attention. Most of the research which exists on the detection and classification of turn lane arrows have many limitations including low accuracy. Therefore, a novel technique based on two novel concepts for improving the performance of the detection and classification of turn lane arrows is proposed in this paper. Firstly, pixel-wise segmentation of all turn lane arrows into one class instead of each turn lane arrow in a separate class is proposed. Secondly, a novel cascaded classifier that evolves its weights so that it can identify turn lane arrows is proposed. Three turn lane road markings named left turn lane, right turn lane and Continuous Central Turning Lane (CCTL) are evaluated using a real-world roadside image dataset created by video data including all state roads in Queensland provided by our industry partners. The comparative analysis of the experimental results demonstrated outstanding results in terms of accuracy.
{"title":"Multi-stage Deep Learning Technique with a Cascaded Classifier for Turn Lanes Recognition","authors":"Pubudu Sanjeewani, B. Verma, J. Affum","doi":"10.1109/SSCI50451.2021.9659973","DOIUrl":"https://doi.org/10.1109/SSCI50451.2021.9659973","url":null,"abstract":"The accurate recognition of road markings such as lanes and turn arrows is required in many applications including autonomous vehicles. Nevertheless, studies on road markings detection are commonly found in literature, detection and classification of turn lane arrows has not gained much attention. Most of the research which exists on the detection and classification of turn lane arrows have many limitations including low accuracy. Therefore, a novel technique based on two novel concepts for improving the performance of the detection and classification of turn lane arrows is proposed in this paper. Firstly, pixel-wise segmentation of all turn lane arrows into one class instead of each turn lane arrow in a separate class is proposed. Secondly, a novel cascaded classifier that evolves its weights so that it can identify turn lane arrows is proposed. Three turn lane road markings named left turn lane, right turn lane and Continuous Central Turning Lane (CCTL) are evaluated using a real-world roadside image dataset created by video data including all state roads in Queensland provided by our industry partners. The comparative analysis of the experimental results demonstrated outstanding results in terms of accuracy.","PeriodicalId":255763,"journal":{"name":"2021 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"8 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":"132694929","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}