Pub Date : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378612
Hamdi Abed, Bálint Gyires-Tóth
Automated machine learning (AutoML) is a technique which helps to determine the optimal or near-optimal model for a specific dataset and has been a focused research area during the last years. The automation of model design opens doors for non-machine learning experts to utilize machine learning models in several scenarios, which is both appealing for a wide range of researchers and for cloud services as well. Neural Architecture Search is a subfield of AutoML where the optimal artificial neural network model's architecture is generally searched with adaptive algorithms. This paper proposes a method to apply Efficient Neural Architecture Search (ENAS) to LSTM-like recurrent architecture, which uses a gating mechanism an inner memory. Using this method, the paper investigates if the handcrafted Long Short-Term Memory (LSTM) cell is an optimal or near-optimal solution of sequence modelling for a given dataset, or other, automatically defined recurrent structures outperform. The performance of vanilla LSTM, and advanced recurrent architectures designed by random search, and reinforcement learning-based ENAS are examined and compared. The proposed methods are evaluated in a text generation task on the Penn TreeBank dataset.
{"title":"Efficient Neural Architecture Search for Long Short-Term Memory Networks","authors":"Hamdi Abed, Bálint Gyires-Tóth","doi":"10.1109/SAMI50585.2021.9378612","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378612","url":null,"abstract":"Automated machine learning (AutoML) is a technique which helps to determine the optimal or near-optimal model for a specific dataset and has been a focused research area during the last years. The automation of model design opens doors for non-machine learning experts to utilize machine learning models in several scenarios, which is both appealing for a wide range of researchers and for cloud services as well. Neural Architecture Search is a subfield of AutoML where the optimal artificial neural network model's architecture is generally searched with adaptive algorithms. This paper proposes a method to apply Efficient Neural Architecture Search (ENAS) to LSTM-like recurrent architecture, which uses a gating mechanism an inner memory. Using this method, the paper investigates if the handcrafted Long Short-Term Memory (LSTM) cell is an optimal or near-optimal solution of sequence modelling for a given dataset, or other, automatically defined recurrent structures outperform. The performance of vanilla LSTM, and advanced recurrent architectures designed by random search, and reinforcement learning-based ENAS are examined and compared. The proposed methods are evaluated in a text generation task on the Penn TreeBank dataset.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133783646","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378622
D. Poștovei, C. Bulac, I. Tristiu, Balduino Estison Mugilila Camachi, N. Anton
The aim of this paper is to improve the automation processes of electrical power substations based on IEC 61850 standard, with the focus on data management related to data structure, path, and communication model. When modelling substations items a simple and clear approach is necessary to effectively match binary or analog data from the switchyard with the IEC 61850 data object model. The study is made by simply transforming data from a single line diagram and its functionalities into object model by explaining its hierarchy structure and how to use it to create an operational process interface. This possible through the three key components of the IEC 61850 standard: Client-Server based on TCP/IP MMS (Manufacturing Messaging Specification) which perform the monitoring and control functions, GOOSE (Generic Object-Oriented Substation Event) protocol used in applications between IEDs (Intelligent Electronic Device) like interlocking signals and trip messages and Sampled Values (SV) protocol used for fast transmission of analogue values over the network.
{"title":"Modelling and implementation of Single Line Diagram data in IEC 61850 environment","authors":"D. Poștovei, C. Bulac, I. Tristiu, Balduino Estison Mugilila Camachi, N. Anton","doi":"10.1109/SAMI50585.2021.9378622","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378622","url":null,"abstract":"The aim of this paper is to improve the automation processes of electrical power substations based on IEC 61850 standard, with the focus on data management related to data structure, path, and communication model. When modelling substations items a simple and clear approach is necessary to effectively match binary or analog data from the switchyard with the IEC 61850 data object model. The study is made by simply transforming data from a single line diagram and its functionalities into object model by explaining its hierarchy structure and how to use it to create an operational process interface. This possible through the three key components of the IEC 61850 standard: Client-Server based on TCP/IP MMS (Manufacturing Messaging Specification) which perform the monitoring and control functions, GOOSE (Generic Object-Oriented Substation Event) protocol used in applications between IEDs (Intelligent Electronic Device) like interlocking signals and trip messages and Sampled Values (SV) protocol used for fast transmission of analogue values over the network.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131161148","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378632
Dipesh Chand, H. Oğul
Increasing demand for lecture videos in digital libraries has raised the challenge of automatic annotation of lecture content for effective navigation of lectures by users. One direction is the prior segmentation of lecture videos to simplify several applications such as indexing, keyword spotting, and targeted search. In this study, we present a lecture video segmentation framework based on the speech content of the instructors. The framework is built upon a model that extracts textual and acoustic features from speech and uses them to identify topical segment boundaries of the lecture video. To evaluate our proposed model, we collected our own dataset containing a diverse set of 37 lecture videos and also manually created ground truth. The performance was measured by using metrics like Precision, Recall, and F1 Score and obtained 0.69, 0.58, and 0.63 respectively. We also compared our model with some previously known similar models where our model outperformed others. The overall results of the study are presented as a lecture video segmentation model, integrating various tools and techniques, and showing promising performance. Findings can be used further for research in content-based search and retrieval using speech content.
{"title":"A Framework for Lecture Video Segmentation from Extracted Speech Content","authors":"Dipesh Chand, H. Oğul","doi":"10.1109/SAMI50585.2021.9378632","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378632","url":null,"abstract":"Increasing demand for lecture videos in digital libraries has raised the challenge of automatic annotation of lecture content for effective navigation of lectures by users. One direction is the prior segmentation of lecture videos to simplify several applications such as indexing, keyword spotting, and targeted search. In this study, we present a lecture video segmentation framework based on the speech content of the instructors. The framework is built upon a model that extracts textual and acoustic features from speech and uses them to identify topical segment boundaries of the lecture video. To evaluate our proposed model, we collected our own dataset containing a diverse set of 37 lecture videos and also manually created ground truth. The performance was measured by using metrics like Precision, Recall, and F1 Score and obtained 0.69, 0.58, and 0.63 respectively. We also compared our model with some previously known similar models where our model outperformed others. The overall results of the study are presented as a lecture video segmentation model, integrating various tools and techniques, and showing promising performance. Findings can be used further for research in content-based search and retrieval using speech content.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134640134","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378658
J. Novotňák, M. Oravec, Jan Hijj, Daniel Jurč
The article deals with the issue of measuring the slip of an asynchronous motor based on the identification of the magnetic field of rotating elements and the rotating field of the stator of an asynchronous motor. By measuring the magnetic field of the rotating elements of the asynchronous motor and the rotating field of the stator, it is possible to identify the corresponding frequencies. Based on this, it is possible to control the slip of the asynchronous motor. This is especially important from the point of view of speed control, reduction of heat losses and from the point of view of increasing the service life of the elements of the frequency converter. Due to the above facts, the measurement was performed on an asynchronous motor with a frequency converter. The measurement results are given in the form of the frequency spectrum of the asynchronous motor and in the form of the value corresponding to the slip of the asynchronous motor. The article also describes the design of a system for controlling the slip of an asynchronous motor.
{"title":"Slip Control by Identifying the Magnetic Field of the Elements of an Asynchronous Motor","authors":"J. Novotňák, M. Oravec, Jan Hijj, Daniel Jurč","doi":"10.1109/SAMI50585.2021.9378658","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378658","url":null,"abstract":"The article deals with the issue of measuring the slip of an asynchronous motor based on the identification of the magnetic field of rotating elements and the rotating field of the stator of an asynchronous motor. By measuring the magnetic field of the rotating elements of the asynchronous motor and the rotating field of the stator, it is possible to identify the corresponding frequencies. Based on this, it is possible to control the slip of the asynchronous motor. This is especially important from the point of view of speed control, reduction of heat losses and from the point of view of increasing the service life of the elements of the frequency converter. Due to the above facts, the measurement was performed on an asynchronous motor with a frequency converter. The measurement results are given in the form of the frequency spectrum of the asynchronous motor and in the form of the value corresponding to the slip of the asynchronous motor. The article also describes the design of a system for controlling the slip of an asynchronous motor.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114081499","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378688
László Szűcs, P. Galambos, D. Drexler
In the past decade, parallel manipulators began gaining more attention since they can outperform their serial counterparts at several areas. The use of common parallel delta robot mechanisms is wide-spread within the industry, especially for fast pick and place applications. There is a new type of parallel mechanism on the rise, called the Generalized Triangle Parallel Robot (GTPR), where the parameters of the robot may differ from arm to arm. Due to the asymmetric structure, the kinematic description of such a mechanism is less trivial. For this reason, we wish to show that many mechanical problems become more straightforward by using screw theory througout the formalism. Screw theory uses the Plücker coordinate representation of mechanical structures. This representation leads to an elegant and tractable form of the kinematics equations. This paper presents a compact tutorial about screw theory and a joint velocity calculation example, with the complete derivation of screws and numerical results.
{"title":"Kinematics of Delta-type Parallel Robot Mechanisms via Screw Theory: A tutorial paper","authors":"László Szűcs, P. Galambos, D. Drexler","doi":"10.1109/SAMI50585.2021.9378688","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378688","url":null,"abstract":"In the past decade, parallel manipulators began gaining more attention since they can outperform their serial counterparts at several areas. The use of common parallel delta robot mechanisms is wide-spread within the industry, especially for fast pick and place applications. There is a new type of parallel mechanism on the rise, called the Generalized Triangle Parallel Robot (GTPR), where the parameters of the robot may differ from arm to arm. Due to the asymmetric structure, the kinematic description of such a mechanism is less trivial. For this reason, we wish to show that many mechanical problems become more straightforward by using screw theory througout the formalism. Screw theory uses the Plücker coordinate representation of mechanical structures. This representation leads to an elegant and tractable form of the kinematics equations. This paper presents a compact tutorial about screw theory and a joint velocity calculation example, with the complete derivation of screws and numerical results.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128094031","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378626
Artem Sokolov, A. Savchenko
This paper is focused on the finetuning of acoustic models for speaker adaptation goals on a given gender. We pretrained the Transformer baseline model on Librispeech-960 and conducted experiments with finetuning on the gender-specific test subsets. The obtained word error rate (WER) relatively to the baseline is up to 5% and 3% lower on male and female subsets, respectively, if the layers in the encoder and decoder are not frozen, and the tuning is started from the last checkpoints. Moreover, we adapted our base model on the complete L2 Arctic dataset of accented speech and finetuned it for particular speakers and male and female genders separately. The models trained on the gender subsets obtained 1–2% lower WER when compared to the model tuned on the whole L2 Arctic dataset. Finally, it was experimentally confirmed that the concatenation of the pretrained voice embeddings (x-vector) and embeddings from a conventional encoder cannot significantly improve the speech recognition accuracy.
{"title":"Gender domain adaptation for automatic speech recognition","authors":"Artem Sokolov, A. Savchenko","doi":"10.1109/SAMI50585.2021.9378626","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378626","url":null,"abstract":"This paper is focused on the finetuning of acoustic models for speaker adaptation goals on a given gender. We pretrained the Transformer baseline model on Librispeech-960 and conducted experiments with finetuning on the gender-specific test subsets. The obtained word error rate (WER) relatively to the baseline is up to 5% and 3% lower on male and female subsets, respectively, if the layers in the encoder and decoder are not frozen, and the tuning is started from the last checkpoints. Moreover, we adapted our base model on the complete L2 Arctic dataset of accented speech and finetuned it for particular speakers and male and female genders separately. The models trained on the gender subsets obtained 1–2% lower WER when compared to the model tuned on the whole L2 Arctic dataset. Finally, it was experimentally confirmed that the concatenation of the pretrained voice embeddings (x-vector) and embeddings from a conventional encoder cannot significantly improve the speech recognition accuracy.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123469030","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378672
Nancy Fulda, Nathaniel R. Robinson
Neural embedding models are often described as having an ‘embedding layer’, or a set of network activations that can be extracted from the model in order to obtain word or sentence representations. In this paper, we show via a modification of the well-known word2vec algorithm that relevant semantic information is contained throughout the entirety of the network, not just in the commonly-extracted hidden layer. This extra information can be extracted by summing embeddings from both the input and output weight matrices of a skip-gram model. Word embeddings generated via this method exhibit strong semantic structure, and are able to outperform traditionally extracted word2vec embeddings in a number of analogy tasks.
{"title":"Improved Word Representations Via Summed Target and Context Embeddings","authors":"Nancy Fulda, Nathaniel R. Robinson","doi":"10.1109/SAMI50585.2021.9378672","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378672","url":null,"abstract":"Neural embedding models are often described as having an ‘embedding layer’, or a set of network activations that can be extracted from the model in order to obtain word or sentence representations. In this paper, we show via a modification of the well-known word2vec algorithm that relevant semantic information is contained throughout the entirety of the network, not just in the commonly-extracted hidden layer. This extra information can be extracted by summing embeddings from both the input and output weight matrices of a skip-gram model. Word embeddings generated via this method exhibit strong semantic structure, and are able to outperform traditionally extracted word2vec embeddings in a number of analogy tasks.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124717258","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378680
Áron Zoltán Kaló, M. Sipos
Optical character recognition systems make it possible to extract text from images. In many cases, this may be sufficient, but there are cases where key-value pairs are required. In this paper, we investigate the use of the open source Tesseract OCR system, to extract text data from images, and perform a key-value pair search. Image noise needs to be minimized with image processing algorithms before recognition. It is necessary to perform so-called post processing procedures on the output of the Tesseract. These post-processors can transform the result of the recognition performed by the OCR system. Those can improve the accuracy of the information extracted during the transformation, for example with the help of regular expressions. The key value pair search is performed after these procedures.
{"title":"Key-Value Pair Searhing System via Tesseract OCR and Post Processing","authors":"Áron Zoltán Kaló, M. Sipos","doi":"10.1109/SAMI50585.2021.9378680","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378680","url":null,"abstract":"Optical character recognition systems make it possible to extract text from images. In many cases, this may be sufficient, but there are cases where key-value pairs are required. In this paper, we investigate the use of the open source Tesseract OCR system, to extract text data from images, and perform a key-value pair search. Image noise needs to be minimized with image processing algorithms before recognition. It is necessary to perform so-called post processing procedures on the output of the Tesseract. These post-processors can transform the result of the recognition performed by the OCR system. Those can improve the accuracy of the information extracted during the transformation, for example with the help of regular expressions. The key value pair search is performed after these procedures.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130088412","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378677
Florian Stalder, Alexander Denzler, L. Mazzola
Nowadays, it is becoming essential to extract knowledge from diverse, large scale data-sources. An effective approach to make knowledge accessible and providing the necessary means for efficient reasoning to take place is through the use of knowledge graphs. The process of building knowledge graphs is usually focused on generating meaningful representations. Hence, applying structure to it, which takes into account the existence of different knowledge domains, their depth and breadth is mostly disregarded. This particular shortcoming leads to a loss of valuable information that could else be harnessed to provide various additional functionalities to an application. In other words, enhancing knowledge graphs in such a way that they can be explored similar to how Google Maps presents the world to us. By zooming in and out, different relevant aspects become visible while unnecessary noise is blended out. Granular computing by itself is more of a theorem that highlights potential benefits from the application of fuzzy and hierarchical structures. Little is said on how a potential granular knowledge graph can be built and which existing clustering algorithms can be used for this task. As such, this paper aims to provide (1) an in-depth view of which critical requirements need to be met by an algorithm to establish a granular structure, (2) the process for how different commonly used algorithms are coping with them, as well as (3) an overview that outlines the different steps in the process of establishing a granular knowledge structure. Two approaches are identified as the most promising ones: for low dimensional data, a Growing Hierarchical Self-Organizing Map (with its adaptive behaviour) and, in case of data with high dimensionality, one approach from the projective clustering family, thanks to their capability of finding strong correlation in sub-spaces of the original dimensions.
{"title":"Towards Granular Knowledge Structures: Comparison of Different Approaches","authors":"Florian Stalder, Alexander Denzler, L. Mazzola","doi":"10.1109/SAMI50585.2021.9378677","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378677","url":null,"abstract":"Nowadays, it is becoming essential to extract knowledge from diverse, large scale data-sources. An effective approach to make knowledge accessible and providing the necessary means for efficient reasoning to take place is through the use of knowledge graphs. The process of building knowledge graphs is usually focused on generating meaningful representations. Hence, applying structure to it, which takes into account the existence of different knowledge domains, their depth and breadth is mostly disregarded. This particular shortcoming leads to a loss of valuable information that could else be harnessed to provide various additional functionalities to an application. In other words, enhancing knowledge graphs in such a way that they can be explored similar to how Google Maps presents the world to us. By zooming in and out, different relevant aspects become visible while unnecessary noise is blended out. Granular computing by itself is more of a theorem that highlights potential benefits from the application of fuzzy and hierarchical structures. Little is said on how a potential granular knowledge graph can be built and which existing clustering algorithms can be used for this task. As such, this paper aims to provide (1) an in-depth view of which critical requirements need to be met by an algorithm to establish a granular structure, (2) the process for how different commonly used algorithms are coping with them, as well as (3) an overview that outlines the different steps in the process of establishing a granular knowledge structure. Two approaches are identified as the most promising ones: for low dimensional data, a Growing Hierarchical Self-Organizing Map (with its adaptive behaviour) and, in case of data with high dimensionality, one approach from the projective clustering family, thanks to their capability of finding strong correlation in sub-spaces of the original dimensions.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"258 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127102100","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 : 2021-01-21DOI: 10.1109/SAMI50585.2021.9378625
Michal Kolárik, M. Sarnovský, Ján Paralič
Data streams can be defined as the continuous stream of data in many forms coming from different sources. Data streams are usually non-stationary with continually changing their underlying structure. Solving of predictive or classification tasks on such data must consider this aspect. Traditional machine learning models applied on the drifting data may become invalid in the case when a concept change appears. To tackle this problem, we must utilize special adaptive learning models, which utilize various tools able to reflect the drifting data. One of the most popular groups of such methods are adaptive ensembles. This paper describes the work focused on the design and implementation of a novel adaptive ensemble learning model, which is based on the construction of a robust ensemble consisting of a heterogeneous set of its members. We used k-NN, Naive Bayes and Hoeffding trees as base learners and implemented an update mechanism, which considers dynamic class-weighting and Q statistics diversity calculation to ensure the diversity of the ensemble. The model was experimentally evaluated on the streaming datasets, and the effects of the diversity calculation were analyzed.
{"title":"Diversity in Ensemble Model for Classification of Data Streams with Concept Drift","authors":"Michal Kolárik, M. Sarnovský, Ján Paralič","doi":"10.1109/SAMI50585.2021.9378625","DOIUrl":"https://doi.org/10.1109/SAMI50585.2021.9378625","url":null,"abstract":"Data streams can be defined as the continuous stream of data in many forms coming from different sources. Data streams are usually non-stationary with continually changing their underlying structure. Solving of predictive or classification tasks on such data must consider this aspect. Traditional machine learning models applied on the drifting data may become invalid in the case when a concept change appears. To tackle this problem, we must utilize special adaptive learning models, which utilize various tools able to reflect the drifting data. One of the most popular groups of such methods are adaptive ensembles. This paper describes the work focused on the design and implementation of a novel adaptive ensemble learning model, which is based on the construction of a robust ensemble consisting of a heterogeneous set of its members. We used k-NN, Naive Bayes and Hoeffding trees as base learners and implemented an update mechanism, which considers dynamic class-weighting and Q statistics diversity calculation to ensure the diversity of the ensemble. The model was experimentally evaluated on the streaming datasets, and the effects of the diversity calculation were analyzed.","PeriodicalId":402414,"journal":{"name":"2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics (SAMI)","volume":"125 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130007240","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：