Unsupervised Feature Learning From Time-Series Data Using Linear Models

Kapourchali, Masoumeh Heidari; Banerjee, Bonny

doi:10.1109/jiot.2018.2845340

Cited by 14 publications

(5 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The computational times for convolution in (12) and the presented model ( 13) are compared at a maximum allowable error of 1% in Fig. 5.…”

Section: A Results Of Dictionary Learningmentioning

confidence: 99%

“…To address this issue, the meaningful features can be learned from time series by performing dictionary learning on highly overlapping time series. Secondly, by comparing their shift-variable versions over different kinds of time series, the signals can be reconstruction, prediction and classification [12]. Previously, several dictionary learning techniques adapted to shift-invariant have been proposed [13]- [15].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Transient Thermal Modeling of Power Semiconductors for Long-Term Load Profiles

Zhang

Zhou

et al. 2023

2023 IEEE Applied Power Electronics Conference and Exposition (APEC)

View full text Add to dashboard Cite

This paper presents an improved thermal analysis method focusing on the long-term profiles of power modules. A shift-invariant dictionary method is developed to extract the critical atoms to sparsely represent time series signals.Then the temperature response of the power module can be approximated by a sparse linear combination of basis functions. The performance of the proposed method is compared with the convolution method, the result indicates that the transient junction temperature of the power module can be predicted with less computational effort. The simulation and experimental results also prove that the proposed method can effectively reconstruct the thermal behavior, exhibits good predictive capability and is suitable for temperature prediction under long time scales.

show abstract

“…The computational times for convolution in (12) and the presented model ( 13) are compared at a maximum allowable error of 1% in Fig. 5.…”

Section: A Results Of Dictionary Learningmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Transient Thermal Modeling of Power Semiconductors for Long-Term Load Profiles

Zhang

Zhou

et al. 2023

2023 IEEE Applied Power Electronics Conference and Exposition (APEC)

View full text Add to dashboard Cite

show abstract

“…Sensors generate high-dimensional multivariate time-series. We use a convolutional sparse coding model (Kapourchali and Banerjee 2018b) to learn a dictionary of features from data. The sequence of indices of the detected feature (o m d ) and corresponding shift (o m τ ) for each variable constitutes the sensory feature vector (the agent's sensory observation):…”

Section: Resultsmentioning

confidence: 99%

EPOC: Efficient Perception via Optimal Communication

Kapourchali

Banerjee

2020

AAAI

Self Cite

View full text Add to dashboard Cite

We propose an agent model capable of actively and selectively communicating with other agents to predict its environmental state efficiently. Selecting whom to communicate with is a challenge when the internal model of other agents is unobservable. Our agent learns a communication policy as a mapping from its belief state to with whom to communicate in an online and unsupervised manner, without any reinforcement. Human activity recognition from multimodal, multisource and heterogeneous sensor data is used as a testbed to evaluate the proposed model where each sensor is assumed to be monitored by an agent. The recognition accuracy on benchmark datasets is comparable to the state-of-the-art even though our model uses significantly fewer parameters and infers the state in a localized manner. The learned policy reduces number of communications. The agent is tolerant to communication failures and can recognize unreliable agents through their communication messages. To the best of our knowledge, this is the first work on learning communication policies by an agent for predicting its environmental state.

show abstract

“…e vast amounts of time-series data generated continuously by smart sensors are essential for the real-time monitoring and intelligent analysis or decision-making of production [1,2]. For example, sudden changes in timeseries data suggest anomalies in the actual production line, and the future trend can be predicted through the periodic pattern of the data [3,4].…”

Section: Introductionmentioning

confidence: 99%

Machine Learning-Based Time-Series Data Analysis in Edge-Cloud-Assisted Oil Industrial IoT System

Shi

Yan

Zhao

et al. 2022

Mobile Information Systems

View full text Add to dashboard Cite

With the rapid development of the Industrial Internet of Things (IIoT) and edge computing techniques, in situ intelligent sensors are continuously generating increasing and vast amounts of time-series data. In many industrial applications, particularly highly distributed systems positioned in remote areas, repeated transmission of large amounts of raw data onto the remote server is not possible. This poses a significant challenge to the timely processing of these data in IIoT. Analyzing and processing all the raw data remotely in the cloud server is impractical and has very low efficiency owing to network latency and the limited cloud computing resources. Failure of detecting abnormal data may result in major production safety problems. Therefore, businesses are moving machine learning capabilities to the edge to enable real-time actions in the field. In this study, we present a machine-learning-based edge-cloud framework to solve this problem. First, robust random cut forest and isolation forest algorithms are employed at the edge gateway to the collected data for the detection of anomalously changing data. Subsequently, these preprocessed time-series data are transmitted to cloud services for data trend prediction and missing data completion using the long short-term memory recurrent neural network method feed in conjunction with the original sequence of historical data combined with the first-order forward difference data. The experimental results show that the machine-learning-based edge-cloud-assisted oil production IIoT system can improve substantially the efficiency and accuracy of time-series data analyses.

show abstract

Unsupervised Feature Learning From Time-Series Data Using Linear Models

Cited by 14 publications

References 28 publications

Transient Thermal Modeling of Power Semiconductors for Long-Term Load Profiles

Transient Thermal Modeling of Power Semiconductors for Long-Term Load Profiles

EPOC: Efficient Perception via Optimal Communication

Machine Learning-Based Time-Series Data Analysis in Edge-Cloud-Assisted Oil Industrial IoT System

Contact Info

Product

Resources

About