Firdaous Essaf, Yujian Li, Seybou Sakho, P. K. Gadosey, Ting Zhang
In lung cancer computer-aided diagnosis (CAD), the correct segmentation of the lung parenchyma is particularly important. In order to reduce the detection area, save computational time and improve accuracy, lung tissue needs to be extracted in advance. An improved method of maximum inter-class variance (OTSU) combined with morphological operations is proposed. First, the original CT image is preprocessed by filtering, denoising, image enhancement, and adaptive threshold binarization; then the connecting area marker obtains the outline, using OTSU-based improvement algorithm to remove interference such as trachea lung fluid, separates the lung essence and background, uses the column scanning, regional color marking and effectively separate the left and right lung leaf adhesion and finally uses a series of morphological operations to repair the extracted lung essence. 830 CT images were selected from the public database LIDC, and were successfully segmented using this proposed method, with an average accuracy of 97.56 percent, an average recall rate that reaches 99.29 percent, and a Dice similarity coefficient of 98.42 percent.
{"title":"An Improved Lung Parenchyma Segmentation Using the Maximum Inter-Class Variance Method (OTSU)","authors":"Firdaous Essaf, Yujian Li, Seybou Sakho, P. K. Gadosey, Ting Zhang","doi":"10.1145/3404555.3404647","DOIUrl":"https://doi.org/10.1145/3404555.3404647","url":null,"abstract":"In lung cancer computer-aided diagnosis (CAD), the correct segmentation of the lung parenchyma is particularly important. In order to reduce the detection area, save computational time and improve accuracy, lung tissue needs to be extracted in advance. An improved method of maximum inter-class variance (OTSU) combined with morphological operations is proposed. First, the original CT image is preprocessed by filtering, denoising, image enhancement, and adaptive threshold binarization; then the connecting area marker obtains the outline, using OTSU-based improvement algorithm to remove interference such as trachea lung fluid, separates the lung essence and background, uses the column scanning, regional color marking and effectively separate the left and right lung leaf adhesion and finally uses a series of morphological operations to repair the extracted lung essence. 830 CT images were selected from the public database LIDC, and were successfully segmented using this proposed method, with an average accuracy of 97.56 percent, an average recall rate that reaches 99.29 percent, and a Dice similarity coefficient of 98.42 percent.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"59 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130869807","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}
In recent years, autonomous driving has become a hot topic, especially in the complex urban road environment. The visual algorithm is the most used scheme for autonomous driving. The traditional condition imitation learning adopts the end-to-end deep learning network. But it lacks interpretability, and the ability of feature extraction and expression of network is limited. There are still some problems in the local planning and detail implementation. To solve these problems, we propose to use the deep residual network architecture and add the dual attention module to learn driving skills, which are closer to human beings. To further improve the detailed feature extraction ability of the network, the deeper residual network architecture is used. To adaptively integrate the global context long-range dependence of the image in the spatial and feature dimensions, the dual attention module is adopted to improve the ability of network expression. At the same time, in order to make full use of the multi-period attribute information of the camera image itself, we redesign the network architecture, extract, integrate the three-way temporal information features and the high-level semantics, and increase the interpretability of the temporal information of the model. This method is tested on the CARLA simulator. The experimental results show that compared with the benchmark algorithm, it achieves better driving effect. Deeper feature extraction and multi-period information fusion can effectively improve the driving ability and driving completion of the agent.
{"title":"Urban Driving Based on Condition Imitation Learning and Multi-Period Information Fusion","authors":"Bolun Ge, Binh Yang, Quan-li Wang","doi":"10.1145/3404555.3404639","DOIUrl":"https://doi.org/10.1145/3404555.3404639","url":null,"abstract":"In recent years, autonomous driving has become a hot topic, especially in the complex urban road environment. The visual algorithm is the most used scheme for autonomous driving. The traditional condition imitation learning adopts the end-to-end deep learning network. But it lacks interpretability, and the ability of feature extraction and expression of network is limited. There are still some problems in the local planning and detail implementation. To solve these problems, we propose to use the deep residual network architecture and add the dual attention module to learn driving skills, which are closer to human beings. To further improve the detailed feature extraction ability of the network, the deeper residual network architecture is used. To adaptively integrate the global context long-range dependence of the image in the spatial and feature dimensions, the dual attention module is adopted to improve the ability of network expression. At the same time, in order to make full use of the multi-period attribute information of the camera image itself, we redesign the network architecture, extract, integrate the three-way temporal information features and the high-level semantics, and increase the interpretability of the temporal information of the model. This method is tested on the CARLA simulator. The experimental results show that compared with the benchmark algorithm, it achieves better driving effect. Deeper feature extraction and multi-period information fusion can effectively improve the driving ability and driving completion of the agent.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133526170","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}
With the advancement of urbanization, urban land use detection has become a research hotspot. Numerous methods have been proposed to identify urban land use, in which points of interest (POI) data is widely used, and sometimes other data source like GPS trajectories is incorporated. However, previous works have hardly fully utilized the global spatial information contained in the POI data, or ignored correlations between features when integrating multiple data source, so resulting in information loss. In this study, we propose an integrated framework titled Region2vec to detect urban land use type by combining POI and mobile phone data. First, POI-based region embeddings are generated by applying Glove model and LDA model to mine the global spatial information and land use topic distributions respectively. The mobile phone data is utilized to generate human activity pattern-based embeddings. Then a similarity matrix is constructed according to POI-based and activity pattern-based embeddings. Finally, the similarity measures are regarded as clustering features to extract the urban land use results. Experiments are implemented and compared with other urban land use algorithms based on data in Sanya, China. The results demonstrate the effectiveness of the proposed framework. This research can provide effective information support for urban planning.
{"title":"Region2vec","authors":"Mingjun Xiang","doi":"10.1145/3404555.3404613","DOIUrl":"https://doi.org/10.1145/3404555.3404613","url":null,"abstract":"With the advancement of urbanization, urban land use detection has become a research hotspot. Numerous methods have been proposed to identify urban land use, in which points of interest (POI) data is widely used, and sometimes other data source like GPS trajectories is incorporated. However, previous works have hardly fully utilized the global spatial information contained in the POI data, or ignored correlations between features when integrating multiple data source, so resulting in information loss. In this study, we propose an integrated framework titled Region2vec to detect urban land use type by combining POI and mobile phone data. First, POI-based region embeddings are generated by applying Glove model and LDA model to mine the global spatial information and land use topic distributions respectively. The mobile phone data is utilized to generate human activity pattern-based embeddings. Then a similarity matrix is constructed according to POI-based and activity pattern-based embeddings. Finally, the similarity measures are regarded as clustering features to extract the urban land use results. Experiments are implemented and compared with other urban land use algorithms based on data in Sanya, China. The results demonstrate the effectiveness of the proposed framework. This research can provide effective information support for urban planning.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"93 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116680934","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}
Yanwu Yang, Xutao Guo, Na Gao, Chenfei Ye, H. T. Ma
In recent years, promising performance of classifying the Alzermerzer's Disease has been achieved by using functional resting-state MRI to extract features by functional connectivity and brain activation in different brain regions such as ReHO, ALFF and so on. However current studies focus on the feature extraction by analyzing the whole time series extracted from the functional images, without considering the variation of the signature changes in the brain regions, which might cause fluctuations of the brain signature activation or the analysis of functional connectivity. This study focus on the image feature automatic encoding and decoding in sequence by a network, where convolutional neural network is used to extract abstract image features in each time step and a long-short term recurrent neural network used to combine features at all time. And finally we use the network to carry out experiments to identify the Alzermerzer's Disease. Our CNN network is developed from the U-net, where we only use the first half of the network to encode the images. Finally we have gained a considerable accuracy in average.
{"title":"An Automatic Encoding and Decoding Method for Differentiating Alzheimer's Disease with Functional MRI","authors":"Yanwu Yang, Xutao Guo, Na Gao, Chenfei Ye, H. T. Ma","doi":"10.1145/3404555.3404570","DOIUrl":"https://doi.org/10.1145/3404555.3404570","url":null,"abstract":"In recent years, promising performance of classifying the Alzermerzer's Disease has been achieved by using functional resting-state MRI to extract features by functional connectivity and brain activation in different brain regions such as ReHO, ALFF and so on. However current studies focus on the feature extraction by analyzing the whole time series extracted from the functional images, without considering the variation of the signature changes in the brain regions, which might cause fluctuations of the brain signature activation or the analysis of functional connectivity. This study focus on the image feature automatic encoding and decoding in sequence by a network, where convolutional neural network is used to extract abstract image features in each time step and a long-short term recurrent neural network used to combine features at all time. And finally we use the network to carry out experiments to identify the Alzermerzer's Disease. Our CNN network is developed from the U-net, where we only use the first half of the network to encode the images. Finally we have gained a considerable accuracy in average.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128944888","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}
In recent years, deep learning has achieved great success in many fields. As the most basic machine learning task, clustering has also become one of the research hotspots. However, clustering performance based on Variational Autoencoder is generally better than that based on Generative Adversarial Network, which is mainly because the former implements multi-modal learning and there are obvious boundaries between different categories, while the latter does not. In this paper, we propose a new clustering model named ACOL-GAN, which replaces the normal distribution that standard GAN relied on with sampling networks and adopts the Auto-clustering Output Layer as the output layer in discriminator. Due to Graph-based Activity Regularization terms, softmax nodes of parent-classes are specialized as the competition between each other during training. The experimental results show that ACOL-GAN achieved the state-of-the-art performance for clustering tasks on MNIST USPS and Fashion-MNIST, with the highest accuracy on Fashion-MNIST.
{"title":"ACOL-GAN: Learning Clustering Generative Adversarial Networks through Graph-Based Activity Regularization","authors":"Songyuan Wu, Liyao Jiao, Qingqiang Wu","doi":"10.1145/3404555.3404581","DOIUrl":"https://doi.org/10.1145/3404555.3404581","url":null,"abstract":"In recent years, deep learning has achieved great success in many fields. As the most basic machine learning task, clustering has also become one of the research hotspots. However, clustering performance based on Variational Autoencoder is generally better than that based on Generative Adversarial Network, which is mainly because the former implements multi-modal learning and there are obvious boundaries between different categories, while the latter does not. In this paper, we propose a new clustering model named ACOL-GAN, which replaces the normal distribution that standard GAN relied on with sampling networks and adopts the Auto-clustering Output Layer as the output layer in discriminator. Due to Graph-based Activity Regularization terms, softmax nodes of parent-classes are specialized as the competition between each other during training. The experimental results show that ACOL-GAN achieved the state-of-the-art performance for clustering tasks on MNIST USPS and Fashion-MNIST, with the highest accuracy on Fashion-MNIST.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128672543","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}
Mariam Bisma, F. Azam, Yawar Rasheed, Muhammad Waseem Anwar
Ensuring security and privacy of IOT devices and the associated/ dependent complex and critical systems is certainly a major concern, especially after proliferation of IoT devices in variety of domains in current era. A considerable level of security can be achieved in these systems using the techniques of Role Based Access Control (RBAC). In contrast to Discretionary Access Control (DAC) where personal identity of the owner/ user matters, RBAC grants access permissions on the basis of roles of the user. Due to the inherent complexity associated with ensuring security in IoT devices and related systems/ services, a level of abstraction is required in the development process, in order to better understand and develop the system accordingly by integrating all the security aspects. This level of abstraction can be achieved by developing the system as per the concepts of Model Driven Development (MDD). In this paper, techniques of Model Driven Architecture (MDA)/ MDD has been used to propose such a Framework/ Meta-Model, which ensures RBAC in order to access the services associated with IoT devices. The proposed Meta-Model can be further extended for the model-based development and automation of such a system that ensure RBAC for IoT devices. Validity of proposed Meta-Model has been proved by creating an M1 level Instance Model of a real-world case study. Results prove, that the proposed Meta-Model is capable to be transformed into a reliable system that ensures RBAC in IoT devices.
{"title":"A Model-Driven Framework for Ensuring Role Based Access Control in IoT Devices","authors":"Mariam Bisma, F. Azam, Yawar Rasheed, Muhammad Waseem Anwar","doi":"10.1145/3404555.3404582","DOIUrl":"https://doi.org/10.1145/3404555.3404582","url":null,"abstract":"Ensuring security and privacy of IOT devices and the associated/ dependent complex and critical systems is certainly a major concern, especially after proliferation of IoT devices in variety of domains in current era. A considerable level of security can be achieved in these systems using the techniques of Role Based Access Control (RBAC). In contrast to Discretionary Access Control (DAC) where personal identity of the owner/ user matters, RBAC grants access permissions on the basis of roles of the user. Due to the inherent complexity associated with ensuring security in IoT devices and related systems/ services, a level of abstraction is required in the development process, in order to better understand and develop the system accordingly by integrating all the security aspects. This level of abstraction can be achieved by developing the system as per the concepts of Model Driven Development (MDD). In this paper, techniques of Model Driven Architecture (MDA)/ MDD has been used to propose such a Framework/ Meta-Model, which ensures RBAC in order to access the services associated with IoT devices. The proposed Meta-Model can be further extended for the model-based development and automation of such a system that ensure RBAC for IoT devices. Validity of proposed Meta-Model has been proved by creating an M1 level Instance Model of a real-world case study. Results prove, that the proposed Meta-Model is capable to be transformed into a reliable system that ensures RBAC in IoT devices.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128412103","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}
Limited battery lifetime and computing ability of size-constrained wireless devices (WDs) have restricted the performance of many low-power wireless networks, e.g. wireless sensor networks and Internet of Things. To that end, our goal is to make a binary offloading policy, so that each computation task of WDs is either executed locally or fully offloaded to a mobile edge computing (MEC) server and further compute the time allocation among multi-users. Specifically, we propose the order-preserving policy generation method, which computationally feasible and efficient in large-size networks and generate various action policies. Then we introduce the bi-section search, using a one-dimensional bisection search over the dual variable associated with the time allocation constraint in O(N) complexity. Finally, extensive simulations show that the proposed methods can effectively achieve a near-optimal performance under various supposed network setups.
{"title":"Meta-Heuristic Search Based Model for Task Offloading and Time Allocation in Mobile Edge Computing","authors":"Yufan Xu, Yan Wang, Junyao Yang","doi":"10.1145/3404555.3404566","DOIUrl":"https://doi.org/10.1145/3404555.3404566","url":null,"abstract":"Limited battery lifetime and computing ability of size-constrained wireless devices (WDs) have restricted the performance of many low-power wireless networks, e.g. wireless sensor networks and Internet of Things. To that end, our goal is to make a binary offloading policy, so that each computation task of WDs is either executed locally or fully offloaded to a mobile edge computing (MEC) server and further compute the time allocation among multi-users. Specifically, we propose the order-preserving policy generation method, which computationally feasible and efficient in large-size networks and generate various action policies. Then we introduce the bi-section search, using a one-dimensional bisection search over the dual variable associated with the time allocation constraint in O(N) complexity. Finally, extensive simulations show that the proposed methods can effectively achieve a near-optimal performance under various supposed network setups.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125247944","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}
P. K. Gadosey, Yuijan Li, Ting Zhang, Zhaoying Liu, Edna Chebet Too, Firdaous Essaf
As a research area of computer vision and deep learning, scene understanding has attracted a lot of attention in recent years. One major challenge encountered is obtaining high levels of segmentation accuracy while dealing with the computational cost and time associated with training or inference. Most current algorithms compromise one metric for the other depending on the intended devices. To address this problem, this paper proposes a novel deep neural network architecture called Segmentation Efficient Blocks Network (SEB-Net) that seeks to achieve the best possible balance between accuracy and computational costs as well as real-time inference speed. The model is composed of both an encoder path and a decoder path in a symmetric structure. The encoder path consists of 16 convolution layers identical to a VGG-19 model, and the decoder path includes what we call E-blocks (Efficient Blocks) inspired by the widely popular ENet architecture's bottleneck module with slight modifications. One advantage of this model is that the max-unpooling in the decoder path is employed for expansion and projection convolutions in the E-Blocks, allowing for less learnable parameters and efficient computation (10.1 frames per second (fps) for a 480x320 input, 11x fewer parameters than DeconvNet, 52.4 GFLOPs for a 640x360 input on a TESLA K40 GPU device). Experimental results on two outdoor scene datasets; Cambridge-driving Labeled Video Database (CamVid) and Cityscapes, indicate that SEB-Net can achieve higher performance compared to Fully Convolutional Networks (FCN), SegNet, DeepLabV, and Dilation8 in most cases. What's more, SEB-Net outperforms efficient architectures like ENet and LinkNet by 16.1 and 11.6 respectively in terms of Instance-level intersection over Union (iLoU). SEB-Net also shows better performance when further evaluated on the SUNRGB-D, an indoor scene dataset
{"title":"SEB-Net: Revisiting Deep Encoder-Decoder Networks for Scene Understanding","authors":"P. K. Gadosey, Yuijan Li, Ting Zhang, Zhaoying Liu, Edna Chebet Too, Firdaous Essaf","doi":"10.1145/3404555.3404629","DOIUrl":"https://doi.org/10.1145/3404555.3404629","url":null,"abstract":"As a research area of computer vision and deep learning, scene understanding has attracted a lot of attention in recent years. One major challenge encountered is obtaining high levels of segmentation accuracy while dealing with the computational cost and time associated with training or inference. Most current algorithms compromise one metric for the other depending on the intended devices. To address this problem, this paper proposes a novel deep neural network architecture called Segmentation Efficient Blocks Network (SEB-Net) that seeks to achieve the best possible balance between accuracy and computational costs as well as real-time inference speed. The model is composed of both an encoder path and a decoder path in a symmetric structure. The encoder path consists of 16 convolution layers identical to a VGG-19 model, and the decoder path includes what we call E-blocks (Efficient Blocks) inspired by the widely popular ENet architecture's bottleneck module with slight modifications. One advantage of this model is that the max-unpooling in the decoder path is employed for expansion and projection convolutions in the E-Blocks, allowing for less learnable parameters and efficient computation (10.1 frames per second (fps) for a 480x320 input, 11x fewer parameters than DeconvNet, 52.4 GFLOPs for a 640x360 input on a TESLA K40 GPU device). Experimental results on two outdoor scene datasets; Cambridge-driving Labeled Video Database (CamVid) and Cityscapes, indicate that SEB-Net can achieve higher performance compared to Fully Convolutional Networks (FCN), SegNet, DeepLabV, and Dilation8 in most cases. What's more, SEB-Net outperforms efficient architectures like ENet and LinkNet by 16.1 and 11.6 respectively in terms of Instance-level intersection over Union (iLoU). SEB-Net also shows better performance when further evaluated on the SUNRGB-D, an indoor scene dataset","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"156 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123197218","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}
Ze-hua Liu, Rui-jie Jiang, Lv-xue Li, Yuyu Zhu, Zheng Mao
This paper presents a solution to improve the evacuation efficiency of the sorting robot and the chances to preserve more assets in an emergency. We propose a danger potential field model for the intelligent sorting warehouse, which takes the number of AGVs between the grid and the exit into account. By taking the danger map calculated by the model as prior knowledge, the paper combines it with Deep Q Network to obtain an effective evacuation scheduling strategy. Finally, comparing the performance of the strategy with the performance of traditional automata and danger potential field in a visual simulator based on the real sorting warehouse using Pygame, the effectiveness and practicability of the model in the paper is verified.
{"title":"Sorting Robots Cluster Evacuation Based on Deep Q Network and Danger Potential Field","authors":"Ze-hua Liu, Rui-jie Jiang, Lv-xue Li, Yuyu Zhu, Zheng Mao","doi":"10.1145/3404555.3404594","DOIUrl":"https://doi.org/10.1145/3404555.3404594","url":null,"abstract":"This paper presents a solution to improve the evacuation efficiency of the sorting robot and the chances to preserve more assets in an emergency. We propose a danger potential field model for the intelligent sorting warehouse, which takes the number of AGVs between the grid and the exit into account. By taking the danger map calculated by the model as prior knowledge, the paper combines it with Deep Q Network to obtain an effective evacuation scheduling strategy. Finally, comparing the performance of the strategy with the performance of traditional automata and danger potential field in a visual simulator based on the real sorting warehouse using Pygame, the effectiveness and practicability of the model in the paper is verified.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129991352","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}
Named Entity Recognition (NER) is a basic task of natural language processing and an indispensable part of machine translation, knowledge mapping and other fields. In this paper, a fusion model of Chinese named entity recognition using BERT, Bidirectional LSTM (BiLSTM) and Conditional Random Field (CRF) is proposed. In this model, Chinese BERT generates word vectors as a word embedding model. Word vectors through BiLSTM can learn the word label distribution. Finally, the model uses Conditional Random Fields to make syntactic restrictions at the sentence level to get annotation sequences. In addition, we can use Whole Word Masking (wwm) instead of the original random mask in BERT's pre-training, which can effectively solve the problem that the word in Chinese NER is partly masked, so as to improve the performance of NER model. In this paper, BERT-wwm (BERT-wwm is the BERT that uses Whole-Word-Masking in pre training tasks), BERT, ELMo and Word2Vec are respectively used for comparative experiments to reflect the effect of bert-wwm in this fusion model. The results show that using Chinese BERT-wwm as the language representation model of NER model has better recognition ability.
命名实体识别(NER)是自然语言处理的一项基本任务，也是机器翻译、知识图谱等领域不可缺少的组成部分。提出了一种基于BERT、双向LSTM (BiLSTM)和条件随机场(CRF)的中文命名实体识别融合模型。在该模型中，中文BERT生成词向量作为词嵌入模型。通过BiLSTM的词向量可以学习到词的标签分布。最后，利用条件随机场在句子层面进行句法限制，得到标注序列。此外，我们可以在BERT的预训练中使用全词掩蔽(Whole Word Masking, wwm)来代替原来的随机掩码，可以有效地解决中文NER中单词部分被掩蔽的问题，从而提高NER模型的性能。本文分别使用BERT-wwm (BERT-wwm是在预训练任务中使用全词掩蔽的BERT)、BERT、ELMo和Word2Vec进行对比实验，以反映BERT-wwm在该融合模型中的效果。结果表明，使用中文BERT-wwm作为NER模型的语言表示模型具有更好的识别能力。
{"title":"Chinese Named Entity Recognition Based on BERT with Whole Word Masking","authors":"Chao Liu, Cui Zhu, Wenjun Zhu","doi":"10.1145/3404555.3404563","DOIUrl":"https://doi.org/10.1145/3404555.3404563","url":null,"abstract":"Named Entity Recognition (NER) is a basic task of natural language processing and an indispensable part of machine translation, knowledge mapping and other fields. In this paper, a fusion model of Chinese named entity recognition using BERT, Bidirectional LSTM (BiLSTM) and Conditional Random Field (CRF) is proposed. In this model, Chinese BERT generates word vectors as a word embedding model. Word vectors through BiLSTM can learn the word label distribution. Finally, the model uses Conditional Random Fields to make syntactic restrictions at the sentence level to get annotation sequences. In addition, we can use Whole Word Masking (wwm) instead of the original random mask in BERT's pre-training, which can effectively solve the problem that the word in Chinese NER is partly masked, so as to improve the performance of NER model. In this paper, BERT-wwm (BERT-wwm is the BERT that uses Whole-Word-Masking in pre training tasks), BERT, ELMo and Word2Vec are respectively used for comparative experiments to reflect the effect of bert-wwm in this fusion model. The results show that using Chinese BERT-wwm as the language representation model of NER model has better recognition ability.","PeriodicalId":220526,"journal":{"name":"Proceedings of the 2020 6th International Conference on Computing and Artificial Intelligence","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131197732","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：