The rapid development of high-throughput sequencing technology has promoted the research of metagenomic sequence. At present, although a large number of sequence classification tools have good classification performance at the genus level and above, there is still room for improvement at the species level. To solve this problem, a metagenomic sequence classification method based on one-dimensional convolutional neural network is proposed in this paper. First, a metagenomic sequence corpus is constructed and used to train word2vec for k-mer embedding. Then, the optimal k value was selected to vectorize the entire gene sequence and serve as the input layer to establish a one-dimensional convolutional neural network classification model to realize species or genus level recognition. Finally, two datasets are used to optimize the model and improve its generalization ability. Experimental results show that the classification performance of this model is almost the same as the genus level, but it improves at the species level and obtains better classification efficiency.
{"title":"Metagenomic Sequence Classification based on One-Dimensional Convolutional Neural Network","authors":"Lei Xiao, Li Deng, Xiao Liu","doi":"10.1145/3581807.3581835","DOIUrl":"https://doi.org/10.1145/3581807.3581835","url":null,"abstract":"The rapid development of high-throughput sequencing technology has promoted the research of metagenomic sequence. At present, although a large number of sequence classification tools have good classification performance at the genus level and above, there is still room for improvement at the species level. To solve this problem, a metagenomic sequence classification method based on one-dimensional convolutional neural network is proposed in this paper. First, a metagenomic sequence corpus is constructed and used to train word2vec for k-mer embedding. Then, the optimal k value was selected to vectorize the entire gene sequence and serve as the input layer to establish a one-dimensional convolutional neural network classification model to realize species or genus level recognition. Finally, two datasets are used to optimize the model and improve its generalization ability. Experimental results show that the classification performance of this model is almost the same as the genus level, but it improves at the species level and obtains better classification efficiency.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121664188","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}
This paper focuses on the cutting edge technology of the Artificial Intelligence, introducing the Decision-Centric Warfare in detail and making it as basis to design the technology framework of auxiliary decision system in marine warfare. On this basis, an application technology framework of intelligent auxiliary decision in marine combat is constructed, which is supported by knowledge driven, data-driven computing model, network training method technology, situation cognition and intelligent decision technology and combat information distributed computing technology. The framework is based on the current successful application examples of artificial intelligence in Deep Mind, Open AI, DARPA and Tencent teams, which are proved practical and operable.
{"title":"From StarCraft II to Military Combat: The Framework of Auxiliary Decision System on Marine Warfare Based on Artificial Intelligence","authors":"Yi Sun, Ju Liu, Qing Sun","doi":"10.1145/3581807.3581895","DOIUrl":"https://doi.org/10.1145/3581807.3581895","url":null,"abstract":"This paper focuses on the cutting edge technology of the Artificial Intelligence, introducing the Decision-Centric Warfare in detail and making it as basis to design the technology framework of auxiliary decision system in marine warfare. On this basis, an application technology framework of intelligent auxiliary decision in marine combat is constructed, which is supported by knowledge driven, data-driven computing model, network training method technology, situation cognition and intelligent decision technology and combat information distributed computing technology. The framework is based on the current successful application examples of artificial intelligence in Deep Mind, Open AI, DARPA and Tencent teams, which are proved practical and operable.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123399874","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}
This paper adopts C++ as a programming tool to develop general algorithms to solve nonlinear algebraic equations. For comparison purpose, five numerical approaches are used. Instead of finding only one root at the neighborhood of a given initial guess, programs are developed in such a way that they can search multiple roots in a specified interval. Several examples are selected to verify the algorithms and their convergence speeds are compared.
{"title":"General Algorithms to Solve Nonlinear Algebraic Equations with C++","authors":"J. Lieh","doi":"10.1145/3581807.3581886","DOIUrl":"https://doi.org/10.1145/3581807.3581886","url":null,"abstract":"This paper adopts C++ as a programming tool to develop general algorithms to solve nonlinear algebraic equations. For comparison purpose, five numerical approaches are used. Instead of finding only one root at the neighborhood of a given initial guess, programs are developed in such a way that they can search multiple roots in a specified interval. Several examples are selected to verify the algorithms and their convergence speeds are compared.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123593695","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}
Facing the increasingly crowded orbital space and the gradually increasing space threats, more attention needs to be paid to spacecraft safety in space. In order to address the problem of non-cooperative spacecraft's approach interference to our spacecraft, the process of non-cooperative spacecraft's approach to our spacecraft by using Hohmann transfer is given by Satellite Tool Kit (STK) software, and the whole process of spacecraft's abnormal orbital maneuvering to approach our spacecraft is identified and judged by the method based on long short-term memory (LSTM) network. The simulation verifies that the LSTM network achieves good results.
{"title":"A Method of Spacecraft Orbit Anomaly Discrimination Based on Long Short-Term Memory Network","authors":"Zonghua Qu, Chunling Wei, Han Yan","doi":"10.1145/3581807.3581891","DOIUrl":"https://doi.org/10.1145/3581807.3581891","url":null,"abstract":"Facing the increasingly crowded orbital space and the gradually increasing space threats, more attention needs to be paid to spacecraft safety in space. In order to address the problem of non-cooperative spacecraft's approach interference to our spacecraft, the process of non-cooperative spacecraft's approach to our spacecraft by using Hohmann transfer is given by Satellite Tool Kit (STK) software, and the whole process of spacecraft's abnormal orbital maneuvering to approach our spacecraft is identified and judged by the method based on long short-term memory (LSTM) network. The simulation verifies that the LSTM network achieves good results.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131285555","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}
The window-based attention is used to alleviate the problem of abrupt increase in computation as the input image resolution grows and shows excellent performance. However, the problem that aggregating global features from different windows is waiting to be resolved. Swin-Transformer is proposed to construct hierarchical encoding by a shifted-window mechanism to interactively learn the information between different windows. In this work, we investigate the outcome of applying an overlapped attention block (MoA) after the local attention layer and apply plenty to medical image segmentation tasks. The overlapped attention module employs slightly larger and overlapped patches in the key and value to enable neighbouring pixel information transmission, which leads to significant performance gain. The experimental results on the ACDC and Synapse datasets demonstrate that the used method performs better than previous Transformer models.
{"title":"MoAFormer: Aggregating Adjacent Window Features into Local Vision Transformer Using Overlapped Attention Mechanism for Volumetric Medical Segmentation","authors":"Yixi Luo, Huayi Yin, X. Du","doi":"10.1145/3581807.3581825","DOIUrl":"https://doi.org/10.1145/3581807.3581825","url":null,"abstract":"The window-based attention is used to alleviate the problem of abrupt increase in computation as the input image resolution grows and shows excellent performance. However, the problem that aggregating global features from different windows is waiting to be resolved. Swin-Transformer is proposed to construct hierarchical encoding by a shifted-window mechanism to interactively learn the information between different windows. In this work, we investigate the outcome of applying an overlapped attention block (MoA) after the local attention layer and apply plenty to medical image segmentation tasks. The overlapped attention module employs slightly larger and overlapped patches in the key and value to enable neighbouring pixel information transmission, which leads to significant performance gain. The experimental results on the ACDC and Synapse datasets demonstrate that the used method performs better than previous Transformer models.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"254 Pt A 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116506529","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}
Renwei Ba, Yidan Zhang, Zhenghui Hu, Jun Sun, Xiao Li
The biological model of the mammal visual mechanisms is very beneficial to feature learning in motionless images. It is proved that the visual mechanisms can improve the performance of the hand-crafted methods and CNNS method. Recently CNNs learn discriminate and robust features by changing the backbone, processing multi-scale feature maps, and adding attention mechanisms, etc. While they are relatively short of changing the network structure with human retina mechanisms, which have been proven to have a strong feature extract capability of images by traditional feature descriptors. To address this problem, we present two CNN blocks, multi-scale receptive field convolutional block (MSRF) and Sensitivity block (SENSI), both of which are constructed by modeling the human retina ganglion cell's mechanisms. MSRF is designed to enhance the feature discriminability and robustness by imitating the exponentially increased way of the receptive fields of the P ganglion cells in the human retina. We constructed experiments to get the specific value of the size of the receptive fields, and it can capture both local and global features with various convolution kernels. SENSI is presented to make sure each receptive field has a suitable weight to choose which receptive field can better learn features. Both of them help to learn features and can be easily integrated into the existing CNN models. The framework is evaluated on two benchmark datasets. We further assemble MSRF and SENSI to the top of SSD, constructing the HRPM Net. The model outperforms the state-of-the-art approaches by a considerable margin on MS COCO, VOC 2012, and VOC 2007 datasets. The results also show that MSRF block and SENSI block are helpful in feature learning and can improve the performance by a margin.
{"title":"HRPM Net: An Efficient Feature Learning Network from The Biological Modelling Of Human Retinal Perception Mechanism","authors":"Renwei Ba, Yidan Zhang, Zhenghui Hu, Jun Sun, Xiao Li","doi":"10.1145/3581807.3581869","DOIUrl":"https://doi.org/10.1145/3581807.3581869","url":null,"abstract":"The biological model of the mammal visual mechanisms is very beneficial to feature learning in motionless images. It is proved that the visual mechanisms can improve the performance of the hand-crafted methods and CNNS method. Recently CNNs learn discriminate and robust features by changing the backbone, processing multi-scale feature maps, and adding attention mechanisms, etc. While they are relatively short of changing the network structure with human retina mechanisms, which have been proven to have a strong feature extract capability of images by traditional feature descriptors. To address this problem, we present two CNN blocks, multi-scale receptive field convolutional block (MSRF) and Sensitivity block (SENSI), both of which are constructed by modeling the human retina ganglion cell's mechanisms. MSRF is designed to enhance the feature discriminability and robustness by imitating the exponentially increased way of the receptive fields of the P ganglion cells in the human retina. We constructed experiments to get the specific value of the size of the receptive fields, and it can capture both local and global features with various convolution kernels. SENSI is presented to make sure each receptive field has a suitable weight to choose which receptive field can better learn features. Both of them help to learn features and can be easily integrated into the existing CNN models. The framework is evaluated on two benchmark datasets. We further assemble MSRF and SENSI to the top of SSD, constructing the HRPM Net. The model outperforms the state-of-the-art approaches by a considerable margin on MS COCO, VOC 2012, and VOC 2007 datasets. The results also show that MSRF block and SENSI block are helpful in feature learning and can improve the performance by a margin.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114976549","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}
Link prediction is a technique to forecast future new or missing relationships between entities based on the current dynamic network information. After a brief introduction of the standard problem and evaluation metrics of link prediction, this review will summarize representative progresses about matrix factorization, probabilistic models, network embedding, deep learning, and some others, mainly extracted from related publications in the last decade. Finally, this review will outline some long-standing challenges for future studies.
{"title":"Progresses in Link Prediction: A Survey","authors":"Jiahao Li, Linlan Liu, Jian Shu","doi":"10.1145/3581807.3581903","DOIUrl":"https://doi.org/10.1145/3581807.3581903","url":null,"abstract":"Link prediction is a technique to forecast future new or missing relationships between entities based on the current dynamic network information. After a brief introduction of the standard problem and evaluation metrics of link prediction, this review will summarize representative progresses about matrix factorization, probabilistic models, network embedding, deep learning, and some others, mainly extracted from related publications in the last decade. Finally, this review will outline some long-standing challenges for future studies.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114482138","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}
Xiang Cheng, Xuan Han, Yu Song, Tielin Zhang, Bo Xu
As brain-related research presents increasing importance, the requirement for automatic spike detection algorithms also emerges. Traditional spike detection algorithms, including amplitude thresholding and wavelet transformation, show several shortcomings that impede the practical application. Here, we propose an artificial neural network-assisted amplitude thresholding algorithm and conduct experiments with raw signals collected from the primary somatosensory cortex and primary motor cortex of macaques. Using F1 score as an evaluation index, artificial neural networks, as well as its lightweight version, effectively help the amplitude thresholding to achieve better performance, showing enormous potential for real-time spike detection application.
{"title":"Artificial Neural Network-assisted Amplitude Thresholding Improves Spike Detection","authors":"Xiang Cheng, Xuan Han, Yu Song, Tielin Zhang, Bo Xu","doi":"10.1145/3581807.3581875","DOIUrl":"https://doi.org/10.1145/3581807.3581875","url":null,"abstract":"As brain-related research presents increasing importance, the requirement for automatic spike detection algorithms also emerges. Traditional spike detection algorithms, including amplitude thresholding and wavelet transformation, show several shortcomings that impede the practical application. Here, we propose an artificial neural network-assisted amplitude thresholding algorithm and conduct experiments with raw signals collected from the primary somatosensory cortex and primary motor cortex of macaques. Using F1 score as an evaluation index, artificial neural networks, as well as its lightweight version, effectively help the amplitude thresholding to achieve better performance, showing enormous potential for real-time spike detection application.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114664093","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}
Hanqi Wang, Huawei Liang, L. Chen, Diancheng Gong, Pengfei Zhou, Bin Kong
Drivable area segmentation is vital for autonomous vehicle driving safety, especially on unstructured roads. Mainstream drivable area algorithms are suited for structured environments, such as urban roads. However, these algorithms perform poorly in unstructured environments. This paper proposes a drivable area segmentation algorithm based on multi-sensor late-fusion for unstructured environments. The algorithm uses the visual segmentation results to correct the light detection and ranging (LiDAR) segmentation results, which can effectively solve those environments with unapparent boundary height differences. Desert experiments show that our algorithm achieves 96.02 on Intersection over Union (IoU), which is 36.75 and 38.31 higher than the LiDAR-based and the Vision-based algorithm, respectively.
可驾驶区域分割对于自动驾驶汽车的驾驶安全至关重要，尤其是在非结构化道路上。主流的可驾驶区域算法适用于结构化环境，如城市道路。然而，这些算法在非结构化环境中表现不佳。提出了一种基于多传感器后期融合的非结构化环境下可驾驶区域分割算法。该算法利用视觉分割结果对激光雷达(LiDAR)分割结果进行校正，可有效解决边界高差不明显的环境问题。沙漠实验结果表明，该算法在Intersection over Union (IoU)上达到96.02，分别比基于lidar的算法和基于vision的算法高36.75和38.31。
{"title":"Drivable Area Segmentation in Unstructured Roads for Autonomous Vehicles based on Multi-sensor Fusion","authors":"Hanqi Wang, Huawei Liang, L. Chen, Diancheng Gong, Pengfei Zhou, Bin Kong","doi":"10.1145/3581807.3581828","DOIUrl":"https://doi.org/10.1145/3581807.3581828","url":null,"abstract":"Drivable area segmentation is vital for autonomous vehicle driving safety, especially on unstructured roads. Mainstream drivable area algorithms are suited for structured environments, such as urban roads. However, these algorithms perform poorly in unstructured environments. This paper proposes a drivable area segmentation algorithm based on multi-sensor late-fusion for unstructured environments. The algorithm uses the visual segmentation results to correct the light detection and ranging (LiDAR) segmentation results, which can effectively solve those environments with unapparent boundary height differences. Desert experiments show that our algorithm achieves 96.02 on Intersection over Union (IoU), which is 36.75 and 38.31 higher than the LiDAR-based and the Vision-based algorithm, respectively.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123538099","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}
Common methods to detect whether a person uses drugs require taking biological samples of the subject, which have time limitation due to the samples. To avoid this, this paper proposes a CNN-based EEG classification method for drug use detection, which does not require taking biological samples of the subject and can trace a longer drug use history of the subject. In this paper, a convolutional neural network-based EEG classification algorithm incorporating batch normalization after the convolutional layer and also introducing dropout operation in the fully connected layer to speed up the training process is designed to distinguish between healthy controls and drug addicts, which reduces the sensitivity of parameters, effectively mitigates the occurrence of overfitting and improves the accuracy compared to traditional machine learning algorithms. Data were collected from eight healthy controls and eight drug addicts. The algorithm obtained the classification accuracy of 85.46% using eight-fold cross-validation. The result of classification shows that the method is an effective way to detect whether the examined person is drug addict, which can easier bring hidden drug addicts under control and reduce the social harm caused by drugs.
{"title":"CNN-based EEG Classification Method for Drug Use Detection","authors":"Hui Zeng, Banghua Yang, Xuelin Gu, Yongcong Li, Xinxing Xia, Shouwei Gao","doi":"10.1145/3581807.3581868","DOIUrl":"https://doi.org/10.1145/3581807.3581868","url":null,"abstract":"Common methods to detect whether a person uses drugs require taking biological samples of the subject, which have time limitation due to the samples. To avoid this, this paper proposes a CNN-based EEG classification method for drug use detection, which does not require taking biological samples of the subject and can trace a longer drug use history of the subject. In this paper, a convolutional neural network-based EEG classification algorithm incorporating batch normalization after the convolutional layer and also introducing dropout operation in the fully connected layer to speed up the training process is designed to distinguish between healthy controls and drug addicts, which reduces the sensitivity of parameters, effectively mitigates the occurrence of overfitting and improves the accuracy compared to traditional machine learning algorithms. Data were collected from eight healthy controls and eight drug addicts. The algorithm obtained the classification accuracy of 85.46% using eight-fold cross-validation. The result of classification shows that the method is an effective way to detect whether the examined person is drug addict, which can easier bring hidden drug addicts under control and reduce the social harm caused by drugs.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"106 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125563219","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：