ISSN 2394-5125
 

Research Article 


INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA

T. Rajalakshmi, S. Sathappan.

Abstract
The world is seeing an exponential growth of Cyber-Physical Systems (CPS), which are projected to
revolutionize the society by developing innovative technologies and applications in a number of fields including
intelligent transport system, environmental monitoring, and mobile health systems. DeepRT is a deep learning technique
used to analysis the CPS data. It focused on supporting predictable temporal and spatial inference performance when
deep learning model are used under unpredictable and resource constrained environments. However, the error rate of
DeepRT is high because of feature space lacking problem. This problem was solved by using Improved Feature Space
Partitioning Tree (IFSPT) where the feature space was split into number of partitions using Classification And
Regression Tree (CART) and the best feature space was selected based on Shapley Feature Importance (SFI) measure.
The selected feature space was processed in Improved DeepRT (IDeepRT) which is the combination of Convolutional
Neural Network (CNN) and Recurrent Neural Network (RNN) to classify the CPS data. In this paper, reinforcement
learning is integrated with IDeepRT to continuously correct feature space representations within IDeepRT. This principle
draws on IDeepRT evidence that reward features during perceptual learning may influence the characteristics of
representations within boundary of classes of complex datasets. Although, the latest high-end computers have powerful
environments, running deeper IDeepRT for complex tasks is challenging. So in this paper, a compressing scheme is
proposed to compress the entire IDeepRT. In the compression scheme, rank selection using bayesian matrix factorization,
tucker decomposition on kernel, fine-tuning are the processes to compress the IDeepRT layers. This scheme provides
significant reductions in model size which effectively reduces the runtime of the model and it also enhances the
classification accuracy.

Key words: Cyber-Physical System, DeepRT, reinforcement learning, compressing scheme, data classification.


 
ARTICLE TOOLS
Abstract
PDF Fulltext
How to cite this articleHow to cite this article
Citation Tools
Related Records
 Articles by T. Rajalakshmi
Articles by S. Sathappan
on Google
on Google Scholar


How to Cite this Article
Pubmed Style

T. Rajalakshmi, S. Sathappan. INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA . JCR. 2020; 7(3): 1366-1375. doi:10.31838/jcr.07.03.214


Web Style

T. Rajalakshmi, S. Sathappan. INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA . http://www.jcreview.com/?mno=102627 [Access: May 31, 2021]. doi:10.31838/jcr.07.03.214


AMA (American Medical Association) Style

T. Rajalakshmi, S. Sathappan. INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA . JCR. 2020; 7(3): 1366-1375. doi:10.31838/jcr.07.03.214



Vancouver/ICMJE Style

T. Rajalakshmi, S. Sathappan. INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA . JCR. (2020), [cited May 31, 2021]; 7(3): 1366-1375. doi:10.31838/jcr.07.03.214



Harvard Style

T. Rajalakshmi, S. Sathappan (2020) INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA . JCR, 7 (3), 1366-1375. doi:10.31838/jcr.07.03.214



Turabian Style

T. Rajalakshmi, S. Sathappan. 2020. INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA . Journal of Critical Reviews, 7 (3), 1366-1375. doi:10.31838/jcr.07.03.214



Chicago Style

T. Rajalakshmi, S. Sathappan. "INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA ." Journal of Critical Reviews 7 (2020), 1366-1375. doi:10.31838/jcr.07.03.214



MLA (The Modern Language Association) Style

T. Rajalakshmi, S. Sathappan. "INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA ." Journal of Critical Reviews 7.3 (2020), 1366-1375. Print. doi:10.31838/jcr.07.03.214



APA (American Psychological Association) Style

T. Rajalakshmi, S. Sathappan (2020) INCORPORATION OF REINFORCEMENT LEARNING AND COMPRESSION OF DEEPRT FOR ACCURATE AND FAST CLASSIFICATION OF CPS DATA . Journal of Critical Reviews, 7 (3), 1366-1375. doi:10.31838/jcr.07.03.214