Pub Date : 2020-12-01DOI: 10.1109/AIKE48582.2020.00030
Satoru Watanabe, H. Yamana
Deep neural networks (DNNs) have improved the performance of artificial intelligence systems in various fields including image analysis, speech recognition, and text classification. However, the consumption of enormous computation resources prevents DNNs from operating on small computers such as edge sensors and handheld devices. Network pruning (NP), which removes parameters from trained DNNs, is one of the prominent methods of reducing the resource consumption of DNNs. In this paper, we propose a novel method of NP, hereafter referred to as PHPM, using persistent homology (PH). PH investigates the inner representation of knowledge in DNNs, and PHPM utilizes the investigation in NP to improve the efficiency of pruning. PHPM prunes DNNs in ascending order of magnitudes of the combinational effects among neurons, which are calculated using the one-dimensional PH, to prevent the deterioration of the accuracy. We compared PHPM with global magnitude pruning method (GMP), which is one of the common baselines to evaluate pruning methods. Evaluation results show that the classification accuracy of DNNs pruned by PHPM outperforms that pruned by GMP.
{"title":"Deep Neural Network Pruning Using Persistent Homology","authors":"Satoru Watanabe, H. Yamana","doi":"10.1109/AIKE48582.2020.00030","DOIUrl":"https://doi.org/10.1109/AIKE48582.2020.00030","url":null,"abstract":"Deep neural networks (DNNs) have improved the performance of artificial intelligence systems in various fields including image analysis, speech recognition, and text classification. However, the consumption of enormous computation resources prevents DNNs from operating on small computers such as edge sensors and handheld devices. Network pruning (NP), which removes parameters from trained DNNs, is one of the prominent methods of reducing the resource consumption of DNNs. In this paper, we propose a novel method of NP, hereafter referred to as PHPM, using persistent homology (PH). PH investigates the inner representation of knowledge in DNNs, and PHPM utilizes the investigation in NP to improve the efficiency of pruning. PHPM prunes DNNs in ascending order of magnitudes of the combinational effects among neurons, which are calculated using the one-dimensional PH, to prevent the deterioration of the accuracy. We compared PHPM with global magnitude pruning method (GMP), which is one of the common baselines to evaluate pruning methods. Evaluation results show that the classification accuracy of DNNs pruned by PHPM outperforms that pruned by GMP.","PeriodicalId":370671,"journal":{"name":"2020 IEEE Third International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126025737","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 : 2020-12-01DOI: 10.1109/AIKE48582.2020.00014
Yuto Tsukagoshi, S. Egami, Y. Sei, Yasuyuki Tahara, Akihiko Ohsuga
For data-driven decision making, it is essential to build a data infrastructure that accumulates various data types. In such organizations as universities, industries, and government bodies, the integration of heterogeneous data and cross-sectional analysis have been an issue as these various data are distributed and stored in different contexts. Knowledge Graphs with a graphical structure that can flexibly change the schema are suitable for such heterogeneous data integration. In this study, we focused on a university campus as an example of a small organization and propose an ontology that enables the cross-sectional analysis of various data. In particular, we semantically interlinked the dimensions in the data model to enable the extraction of data across multiple domains from various perspectives. Then, the unstructured data collected were accumulated as knowledge Graphs based on the proposed ontology to build a data infrastructure. In addition, we found several correlations that could help in solving university campus issues and improving university management using the developed ontology-based data infrastructure.
{"title":"Ontology-Based Correlation Detection Among Heterogeneous Data Sets: A Case Study of University Campus Issues","authors":"Yuto Tsukagoshi, S. Egami, Y. Sei, Yasuyuki Tahara, Akihiko Ohsuga","doi":"10.1109/AIKE48582.2020.00014","DOIUrl":"https://doi.org/10.1109/AIKE48582.2020.00014","url":null,"abstract":"For data-driven decision making, it is essential to build a data infrastructure that accumulates various data types. In such organizations as universities, industries, and government bodies, the integration of heterogeneous data and cross-sectional analysis have been an issue as these various data are distributed and stored in different contexts. Knowledge Graphs with a graphical structure that can flexibly change the schema are suitable for such heterogeneous data integration. In this study, we focused on a university campus as an example of a small organization and propose an ontology that enables the cross-sectional analysis of various data. In particular, we semantically interlinked the dimensions in the data model to enable the extraction of data across multiple domains from various perspectives. Then, the unstructured data collected were accumulated as knowledge Graphs based on the proposed ontology to build a data infrastructure. In addition, we found several correlations that could help in solving university campus issues and improving university management using the developed ontology-based data infrastructure.","PeriodicalId":370671,"journal":{"name":"2020 IEEE Third International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123767008","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 : 2020-12-01DOI: 10.1109/AIKE48582.2020.00018
A. Nesen, B. Bhargava
Video understanding, surveillance and analytics fields have been dynamically expanding over the recent years due to the enormous amount of CCTV, dashcams and phone cameras which generate video data stored on cloud servers, in social networks, in public and private repositories. The video data has a great potential to be used for improving situation awareness, prediction and prevention of unwanted events and disasters in various settings. Still, there is a significant need for methods and ways to understand the large amount of video recordings and to extract hidden patterns and knowledge. Deep learning networks have been successfully applied for video object and anomaly detection tasks. However, while neural networks focus on utilizing features within an object to be detected, the vast amount of background knowledge remains unnoticed. We propose a semantics centered method for video anomaly detection which allows to identify entities that are inconsistent with the scene and thus can be marked as a potential anomaly. Our method is inspired with the way humans comprehend the surroundings with incorporating external knowledge and previous experience. As a source of external knowledge for deep learning networks we maintain a knowledge graph which allows to compute semantic similarity between the detected objects. Similarity of the entities in the frame depends on the distance between the graph vertices which represent the recognized entities. The object which is semantically distinct from other entities in the video is an anomalous one. We conduct experiments on real-life data to empirically prove the efficiency of our approach and provide an enhanced framework that leads to anomaly detection in video with higher accuracy and better interpretability.
{"title":"Knowledge Graphs for Semantic-Aware Anomaly Detection in Video","authors":"A. Nesen, B. Bhargava","doi":"10.1109/AIKE48582.2020.00018","DOIUrl":"https://doi.org/10.1109/AIKE48582.2020.00018","url":null,"abstract":"Video understanding, surveillance and analytics fields have been dynamically expanding over the recent years due to the enormous amount of CCTV, dashcams and phone cameras which generate video data stored on cloud servers, in social networks, in public and private repositories. The video data has a great potential to be used for improving situation awareness, prediction and prevention of unwanted events and disasters in various settings. Still, there is a significant need for methods and ways to understand the large amount of video recordings and to extract hidden patterns and knowledge. Deep learning networks have been successfully applied for video object and anomaly detection tasks. However, while neural networks focus on utilizing features within an object to be detected, the vast amount of background knowledge remains unnoticed. We propose a semantics centered method for video anomaly detection which allows to identify entities that are inconsistent with the scene and thus can be marked as a potential anomaly. Our method is inspired with the way humans comprehend the surroundings with incorporating external knowledge and previous experience. As a source of external knowledge for deep learning networks we maintain a knowledge graph which allows to compute semantic similarity between the detected objects. Similarity of the entities in the frame depends on the distance between the graph vertices which represent the recognized entities. The object which is semantically distinct from other entities in the video is an anomalous one. We conduct experiments on real-life data to empirically prove the efficiency of our approach and provide an enhanced framework that leads to anomaly detection in video with higher accuracy and better interpretability.","PeriodicalId":370671,"journal":{"name":"2020 IEEE Third International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128379289","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 : 2020-12-01DOI: 10.1109/AIKE48582.2020.00019
Ying-Jia Lin, Daniel Tan, Tzu-Hsuan Chou, Hung-Yu kao, Hsin-Yang Wang
Large-scale pre-trained frameworks have shown state-of-the-art performance in several natural language processing tasks. However, the costly training and inference time are great challenges when deploying such models to real-world applications. In this work, we conduct an empirical study of knowledge distillation on an extractive text summarization task. We first utilized a pre-trained model as the teacher model for extractive summarization and extracted learned knowledge from it as soft targets. Then, we leveraged both the hard targets and the soft targets as the objective for training a much smaller student model to perform extractive summarization. Our results show the student model performs only 1 point lower in the three ROUGE scores on the CNN/DM dataset of extractive summarization while being 40% smaller than the teacher model and 50% faster in terms of the inference time.
{"title":"Knowledge Distillation on Extractive Summarization","authors":"Ying-Jia Lin, Daniel Tan, Tzu-Hsuan Chou, Hung-Yu kao, Hsin-Yang Wang","doi":"10.1109/AIKE48582.2020.00019","DOIUrl":"https://doi.org/10.1109/AIKE48582.2020.00019","url":null,"abstract":"Large-scale pre-trained frameworks have shown state-of-the-art performance in several natural language processing tasks. However, the costly training and inference time are great challenges when deploying such models to real-world applications. In this work, we conduct an empirical study of knowledge distillation on an extractive text summarization task. We first utilized a pre-trained model as the teacher model for extractive summarization and extracted learned knowledge from it as soft targets. Then, we leveraged both the hard targets and the soft targets as the objective for training a much smaller student model to perform extractive summarization. Our results show the student model performs only 1 point lower in the three ROUGE scores on the CNN/DM dataset of extractive summarization while being 40% smaller than the teacher model and 50% faster in terms of the inference time.","PeriodicalId":370671,"journal":{"name":"2020 IEEE Third International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128471342","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 : 2020-12-01DOI: 10.1109/AIKE48582.2020.00016
A. Periola, A. Alonge, K. Ogudo
Cognitive radios (CRs) use artificial intelligence algorithms to obtain an improved quality of service (QoS). CRs also benefit from meta—cognition algorithms that enable them to determine the most suitable intelligent algorithm for achieving their operational goals. Examples of intelligent algorithms that are used by CRs are support vector machines, artificial neural networks and hidden markov models. Each of these intelligent algorithms can be realized in a different manner and used for different tasks such as predicting the idle state and duration of a channel. The CR benefits from jointly using these intelligent algorithms and selecting the most suitable algorithm for prediction at an epoch of interest. The incorporation of meta-cognition also furnishes the CR with consciousness. This is because it makes the CR aware of its learning mechanisms. CR consciousness consumes the CR resources i.e. battery and memory. The resource consumption should be reduced to enhance CR's resources available for data transmission. The discussion in this paper proposes a meta—cognitive solution that reduces CR resources associated with maintaining consciousness. The proposed solution incorporates the time domain and uses information on the duration associated with executing learning and data transmission tasks. In addition, the proposed solution is integrated in a multimode CR. Evaluation shows that the performance improvement for the CR transceiver power, computational resources and channel capacity lies in the range 18.3% – 42.5% , 21.6% – 44.8% and 9.5% – 56.3% on average, respectively.
{"title":"Architecture Model for Wireless Network Conscious Agent","authors":"A. Periola, A. Alonge, K. Ogudo","doi":"10.1109/AIKE48582.2020.00016","DOIUrl":"https://doi.org/10.1109/AIKE48582.2020.00016","url":null,"abstract":"Cognitive radios (CRs) use artificial intelligence algorithms to obtain an improved quality of service (QoS). CRs also benefit from meta—cognition algorithms that enable them to determine the most suitable intelligent algorithm for achieving their operational goals. Examples of intelligent algorithms that are used by CRs are support vector machines, artificial neural networks and hidden markov models. Each of these intelligent algorithms can be realized in a different manner and used for different tasks such as predicting the idle state and duration of a channel. The CR benefits from jointly using these intelligent algorithms and selecting the most suitable algorithm for prediction at an epoch of interest. The incorporation of meta-cognition also furnishes the CR with consciousness. This is because it makes the CR aware of its learning mechanisms. CR consciousness consumes the CR resources i.e. battery and memory. The resource consumption should be reduced to enhance CR's resources available for data transmission. The discussion in this paper proposes a meta—cognitive solution that reduces CR resources associated with maintaining consciousness. The proposed solution incorporates the time domain and uses information on the duration associated with executing learning and data transmission tasks. In addition, the proposed solution is integrated in a multimode CR. Evaluation shows that the performance improvement for the CR transceiver power, computational resources and channel capacity lies in the range 18.3% – 42.5% , 21.6% – 44.8% and 9.5% – 56.3% on average, respectively.","PeriodicalId":370671,"journal":{"name":"2020 IEEE Third International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129232149","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 : 2020-10-01DOI: 10.1109/dsaa49011.2020.00006
{"title":"Message from Program Chairs","authors":"","doi":"10.1109/dsaa49011.2020.00006","DOIUrl":"https://doi.org/10.1109/dsaa49011.2020.00006","url":null,"abstract":"","PeriodicalId":370671,"journal":{"name":"2020 IEEE Third International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124597651","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}
{"title":"Title Page iii","authors":"","doi":"10.1109/sc2.2018.00002","DOIUrl":"https://doi.org/10.1109/sc2.2018.00002","url":null,"abstract":"","PeriodicalId":370671,"journal":{"name":"2020 IEEE Third International Conference on Artificial Intelligence and Knowledge Engineering (AIKE)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134428943","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：