Multivariate time series is ubiquitous in real-world applications, yet it often suffers from missing values that impede downstream analytical tasks. In this paper, we introduce the Long Short-Term Memory Network based...
详细信息
Multivariate time series is ubiquitous in real-world applications, yet it often suffers from missing values that impede downstream analytical tasks. In this paper, we introduce the Long Short-Term Memory Network based recurrent autoencoder with Imputation Units and Temporal Attention Imputation Model (RATAI), tailored for multivariate time series. RATAI is designed to address certain limitations of traditional RNN-based imputation methods, which often focus on predictive modeling to estimate missing values, sometimes neglecting the contextual impact of observed data at and beyond the target time step. Drawing inspiration from Kalman smoothing, which effectively integrates past and future information to refine state estimations, RATAI aims to extract feature representations from time series data and use them to reconstruct a complete time series, thus overcoming the shortcomings of existing approaches. It employs a dual-stage imputation process: the encoder utilizes temporal information and attribute correlations to predict and impute missing values, and extract feature representation of imputed time series. Subsequently, the decoder reconstructs the series from the feature representation, and the reconstructed values are used as the final imputation values. Additionally, RATAI incorporates a temporal attention mechanism, allowing the decoder to focus on highly relevant inputs during reconstruction. This model can be trained directly using data that contains missing values, avoiding the misleading effects on model training that can arise from setting initial values for missing values. Our experiments demonstrate that RATAI outperforms benchmark models in multivariate time series imputation.
Leveraging the information-rich and large volume of Electronic Health Records (EHR), deep learning systems have shown great promise in assisting medical diagnosis and regulatory decisions. Although deep learning model...
详细信息
ISBN:
(纸本)9783030676674;9783030676667
Leveraging the information-rich and large volume of Electronic Health Records (EHR), deep learning systems have shown great promise in assisting medical diagnosis and regulatory decisions. Although deep learning models have advantages over the traditional machine learning approaches in the medical domain, the discovery of adversarial examples has exposed great threats to the state-of-art deep learning medical systems. While most of the existing studies are focused on the impact of adversarial perturbation on medical images, few works have studied adversarial examples and potential defenses on temporal EHR data. In this work, we propose RADAR, a recurrent autoencoder based Detector for Adversarial examples on temporal EHR data, which is the first effort to defend adversarial examples on temporal EHR data. We evaluate RADAR on a mortality classifier using the MIMIC-III dataset. Experiments show that RADAR can filter out more than 90% of adversarial examples and improve the target model accuracy by more than 90% and F1 score by 60%. Besides, we also propose an enhanced attack by introducing the distribution divergence into the loss function such that the adversarial examples are more realistic and difficult to detect.
Multi-view graph clustering aims to discover communities or groups in the graph with multiple views, which usually can supply more comprehensive information than that single-view graph clustering. With the increasing ...
详细信息
ISBN:
(纸本)9781450385053
Multi-view graph clustering aims to discover communities or groups in the graph with multiple views, which usually can supply more comprehensive information than that single-view graph clustering. With the increasing scale of complex data from the real world, multi-view graph clustering has drawn much attention. It has a solid theoretical foundation and high effectiveness in applications such as data mining and social network analysis. However, Most existing methods obtain the clustering result only through the shared feature representations, defectively overlooking the unique features of multiple views. To fill this gap, a Graph recurrent autoencoder (GRAE) is proposed for attributed multi-view graph clustering, which can attain node representation well by learning different view features. Specifically, we first design a global graph autoencoder and a partial graph autoencoder to extract the shared features and the unique features of all views, respectively, which can better represent the nodes in the graph. Then, from the perspective of representation fusion, we adopt an adaptive weight learning method to fuse the different features according to the importance of features. Moreover, we investigate a self-training clustering method to optimize a clustering objective for improving the clustering effect. Finally, we conducted a large number of experiments on three real-world datasets, demonstrating the superior performance of our proposed GRAE model on the multi-view graph clustering task.
In this work, we propose a new recurrent autoencoder architecture, termed Feedback recurrent autoencoder (FRAE), for online compression of sequential data with temporal dependency. The recurrent structure of FRAE is d...
详细信息
ISBN:
(纸本)9781509066315
In this work, we propose a new recurrent autoencoder architecture, termed Feedback recurrent autoencoder (FRAE), for online compression of sequential data with temporal dependency. The recurrent structure of FRAE is designed to efficiently extract the redundancy along the time dimension and allows a compact discrete representation of the data to be learned. We demonstrate its effectiveness in speech spectrogram compression. Specifically, we show that the FRAE, paired with a powerful neural vocoder, can produce high-quality speech waveforms at a low, fixed bitrate. We further show that by adding a learned prior for the latent space and using an entropy coder, we can achieve an even lower variable bitrate.
The increasing complexity of mobility plus the growing population in cities, together with the importance of privacy when sharing data from vehicles or any device, makes traffic forecasting that uses data from infrast...
详细信息
The increasing complexity of mobility plus the growing population in cities, together with the importance of privacy when sharing data from vehicles or any device, makes traffic forecasting that uses data from infrastructure and citizens an open and challenging task. In this paper, we introduce a novel approach to deal with predictions of volume, speed and main traffic direction, in a new aggregated way of traffic data presented as videos. Our approach leverages the continuity in a sequence of frames, learning to embed them into a low dimensional space with an encoder and making predictions there using recurrent layers, ensuring good performance through an embedded loss, and then, recovering back spatial dimensions with a decoder using a second loss at a pixel level. Exogenous variables like weather, time and calendar are also added in the model. Furthermore, we introduce a novel sampling approach for sequences that ensures diversity when creating batches, running in parallel to the optimization process.
Multi-view graph clustering aims to discover communities or groups in the graph with multiple views, which usually can supply more comprehensive information than that single-view graph clustering. With the increasing ...
详细信息
Multi-view graph clustering aims to discover communities or groups in the graph with multiple views, which usually can supply more comprehensive information than that single-view graph clustering. With the increasing scale of complex data from the real world,multi-view graph clustering has drawn much attention. It has a solid theoretical foundation and high effectiveness in applications such as data mining and social network analysis. However, Most existing methods obtain the clustering result only through the shared feature representations, defectively overlooking the unique features of multiple views. To fill this gap, a Graph recurrent Auto Encoder(GRAE)is proposed for attributed multi-view graph clustering, which can attain node representation well by learning different view ***, we first design a global graph autoencoder and a partial graph autoencoder to extract the shared features and the unique features of all views, respectively, which can better represent the nodes in the graph. Then, from the perspective of representation fusion, we adopt an adaptive weight learning method to fuse the different features according to the importance of features. Moreover, we investigate a self-training clustering method to optimize a clustering objective for improving the clustering effect. Finally,we conducted a large number of experiments on three real-world datasets, demonstrating the superior performance of our proposed GRAE model on the multi-view graph clustering task.
Trend fuzzy information granulation has shown promising results in long-term time-series forecasting and has attracted increasing attention. In the forecasting model based on trend fuzzy information granulation, the r...
详细信息
Trend fuzzy information granulation has shown promising results in long-term time-series forecasting and has attracted increasing attention. In the forecasting model based on trend fuzzy information granulation, the representation of trend granules plays a crucial role. The research focuses on developing trend granules and trend granular time series to effectively represent trend information and improve forecasting performance. However, the existing trend fuzzy information granulation methods make assumptions about the trend pattern of granules (i.e., assuming that granules have linear trends or definite nonlinear trends). Fuzzy information granules with presupposed trend patterns have limited expressive ability and struggle to capture complex nonlinear trends and temporal dependencies, thus limiting their forecasting performance. To address this issue, this paper proposes a novel kind of trend fuzzy information granules, named Trend-Pattern Unlimited Fuzzy Information Granules (TPUFIGs), which are constructed by the recurrent autoencoder with automatic feature learning and nonlinear modeling capabilities. Compared with the existing trend fuzzy information granules, TPUFIGs can better characterize potential trend patterns and temporal dependencies, and exhibit stronger robustness. With the TPUFIGs and Long Short-Term Memory (LSTM) neural network, we design the TPUFIG-LSTM forecasting model, which can effectively alleviate error accumulation and improve forecasting capability. Experimental results on six heterogeneous time series datasets demonstrate the superior performance of the proposed model. By combining deep learning and granular computing, this fuzzy information granulation method characterizes intricate dynamic features in time series more effectively, thus providing a novel solution for long-term time series forecasting with improved forecasting accuracy and generalization capability.
Objective: The wearable and portable Electroencephalogram (EEG) sensing systems are deeply interfered by unavoidable physiological artifacts due to the limited recording resources. In this work, an intelligent artifac...
详细信息
Objective: The wearable and portable Electroencephalogram (EEG) sensing systems are deeply interfered by unavoidable physiological artifacts due to the limited recording resources. In this work, an intelligent artifact removal system that handles single-channel EEG signals in the presence of mixed multi-type artifacts is investigated. Methods: The basic idea is to represent the mixed artifacts in contaminated varying EEG signals with the unchanged latent pattern features, and then employ the adaptive artifact removal scheme to separate the contamination and clean EEG signals in the encoded feature domain. To minimize the risks of corrupting clean signals and keeping artifacts by mistake, the artifact removal is formulated as an identification-removal two-stage minimization problem, and an attention based adaptive feature concentration mechanism is designed to improve the removal utility and reduce the calculation consumption. Results: In the real implementation on open real-world dataset, this study achieves the artifact identification accuracy of 98.52% and average correlation coefficient of 0.73 for the removal of strong mixed multi-type artifacts. Conclusion: This study can deal with single-channel EEG signals contaminated by mixed multi-type artifacts with high accuracy and low overhead, and is more effective and stable than traditional schemes with fixed criteria. Significance: This study can significantly improve the signal quality acquired by simplified EEG sensing systems, and may extend the application of wearable and portable EEG sensing systems to medical diagnosis, cognitive science research and other applications requiring clinical setups.
Acoustic Event Classification (AEC) has become a significant task for machines to perceive the surrounding auditory scene. However, extracting effective representations that capture the underlying characteristics of t...
详细信息
Acoustic Event Classification (AEC) has become a significant task for machines to perceive the surrounding auditory scene. However, extracting effective representations that capture the underlying characteristics of the acoustic events is still challenging. Previous methods mainly focused on designing the audio features in a 'handcrafted' manner. Interestingly, data-learnt features have been recently reported to show better performance. Up to now, these were only considered on the frame-level. In this article, we propose an unsupervised learning framework to learn a vector representation of an audio sequence for AEC. This framework consists of a recurrent Neural Network (RNN) encoder and a RNN decoder, which respectively transforms the variable-length audio sequence into a fixed-length vector and reconstructs the input sequence on the generated vector. After training the encoder-decoder, we feed the audio sequences to the encoder and then take the learnt vectors as the audio sequence representations. Compared with previous methods, the proposed method can not only deal with the problem of arbitrary-lengths of audio streams, but also learn the salient information of the sequence. Extensive evaluation on a large-size acoustic event database is performed, and the empirical results demonstrate that the learnt audio sequence representation yields a significant performance improvement by a large margin compared with other state-of-the-art hand-crafted sequence features for AEC.
The paper introduces a dynamic extension of the hybrid random field (HRF), called dynamic HRF (D-HRF). The D-HRF is aimed at the probabilistic graphical modeling of arbitrary-length sequences of sets of (time-dependen...
详细信息
The paper introduces a dynamic extension of the hybrid random field (HRF), called dynamic HRF (D-HRF). The D-HRF is aimed at the probabilistic graphical modeling of arbitrary-length sequences of sets of (time-dependent) discrete random variables under Markov assumptions. Suitable maximum likelihood algorithms for learning the parameters and the structure of the D-HRF are presented. The D-HRF inherits the computational efficiency and the modeling capabilities of HRFs, subsuming both dynamic Bayesian networks and Markov random fields. The behavior of the D-HRF is first evaluated empirically on synthetic data drawn from probabilistic distributions having known form. Then, D-HRFs (combined with a recurrent autoencoder) are successfully applied to the prediction of the disulfide-bonding state of cysteines from the primary structure of proteins in the Protein Data Bank.
暂无评论