Computational capabilities of recurrent NARX neural networks

Siegelmann, Hava T.; Horne, Bill G.; Giles, C. Lee

doi:10.1109/3477.558801

Cited by 379 publications

(180 citation statements)

References 24 publications

Supporting

Mentioning

171

Contrasting

Unclassified

Order By: Relevance

“…NARXnns have been shown to perform well on problems involving long-term dependencies, and they are capable of simulating universal dynamical systems [52,53]. The feasibility of NARXnns as a nonlinear tool for time series modeling and prediction is key for carrying out long-term time series predictions [54].…”

Section: Narxnn and Its Configuration For Sm Retrievalmentioning

confidence: 99%

Estimating Time Series Soil Moisture by Applying Recurrent Nonlinear Autoregressive Neural Networks to Passive Microwave Data over the Heihe River Basin, China

et al. 2017

View full text Add to dashboard Cite

Abstract:A method using a nonlinear auto-regressive neural network with exogenous input (NARXnn) to retrieve time series soil moisture (SM) that is spatially and temporally continuous and high quality over the Heihe River Basin (HRB) in China was investigated in this study. The input training data consisted of the X-band dual polarization brightness temperature (TB) and the Ka-band V polarization TB from the Advanced Microwave Scanning Radiometer II (AMSR2), Global Land Satellite product (GLASS) Leaf Area Index (LAI), precipitation from the Tropical Rainfall Measuring Mission (TRMM) and the Global Precipitation Measurement (GPM), and a global 30 arc-second elevation (GTOPO-30). The output training data were generated from fused SM products of the Japan Aerospace Exploration Agency (JAXA) and the Land Surface Parameter Model (LPRM). The reprocessed fused SM from two years (2013 and 2014) was inputted into the NARXnn for training; subsequently, SM during a third year (2015) was estimated. Direct and indirect validations were then performed during the period 2015 by comparing with in situ measurements, SM from JAXA, LPRM and the Global Land Data Assimilation System (GLDAS), as well as precipitation data from TRMM and GPM. The results showed that the SM predictions from NARXnn performed best, as indicated by their higher correlation coefficients (R ≥ 0.85 for the whole year of 2015), lower Bias values (absolute value of Bias ≤ 0.02) and root mean square error values (RMSE ≤ 0.06), and their improved response to precipitation. This method is being used to produce the NARXnn SM product over the HRB in China.

show abstract

Section: Narxnn and Its Configuration For Sm Retrievalmentioning

confidence: 99%

Estimating Time Series Soil Moisture by Applying Recurrent Nonlinear Autoregressive Neural Networks to Passive Microwave Data over the Heihe River Basin, China

et al. 2017

View full text Add to dashboard Cite

show abstract

“…In theory, it has been shown that the NARX networks can be used, rather than conventional recurrent networks, without computational loss and that they are equivalent to Turing machines [27]. In this case, u(t) and y(t) represent the input and the output of the network at time t, respectively.…”

Section: Recurrent Narx Neural Networkmentioning

confidence: 99%

Hybrid Neural Network Predictive-Wavelet Image Compression System

et al. 2015

View full text Add to dashboard Cite

This paper considers a novel image compression technique called Hybrid Predictive Wavelet coding. The new proposed technique combines the properties of predictive coding and discrete Wavelet coding. In contrast to JPEG2000, the image data values are pre-processed using predictive coding to remove inter-pixel redundancy. The error values, which are the difference between the original and the predicted values, are discrete wavelet coding transformed. In this case, a nonlinear neural network predictor is utilised in the predictive coding system. The simulation results indicated that the proposed technique can achieve good compressed images at high decomposition levels in comparison to JPEG2000.The authors would like to thanks the reviewers for their constructive comments which have certainly improved the quality of the paper significantly.Reviewer #1: -The paper proposes an image compression system based on neural networks. I found the paper well written, technically sound and with clear focus.-The authors have provided a good study of related techniques and the proposed approach is well motivated. I have two minor comments to improve the experimental part of the paper further.The authors would like to thanks the reviewer for his/her encouraging comment.-The first comment concerns the quantitative results. I suggest that the authors study the statistical significance of the results as compared to other approaches. Actually it will be helpful to test if the difference between the proposed approach and comparable approaches is statistically significant.To check the statistical significant between the proposed HNNPWA and JPPEG 2000 techniques, the authors performed a paired t-test based on the Absolute value of the error image. The calculated tvalues showed that the proposed technique outperform JPEG2000 with α = 5% significance level for a one-tailed test at decomposition levels 4, 5 and 6. The t-test indicated that there is no significant difference between the two image compression techniques at decomposition level 3. As it can be noted from Figure 8, at decomposition levels 1, 2 and 3, the visual quality of the reconstructed images for the proposed HPNNWA and JPEG2000 is very good and it is not easy to notice the difference between the original image and the constructed image for both systems.-The comparison is mainly based on the PSNR which is the most widely used approach. Are there other measures that can be used to emphasize further the advantages of the proposed approach.For all the experiments, the authors have added the mean absolute value of the error as another quality measure.-Finally, the conclusion could be improved further by adding more analysis and discussions that could be devoted to explaining the main problems that could be related to the application of the proposed approach.The conclusion section was expanded and the problem of the application of the proposed system was mentioned as requested by the reviewer.Reviewer #2: The quality of written presentation demonstrates a good standard ...

show abstract

“…Several recent papers have considered the nature of deterministic analog (continuous) recurrent neural networks (e.g., [41,43,42,25,22,28,14,44]). These consist of a finite number of neurons.…”

Section: Deterministic Analog Recurrent Networkmentioning

confidence: 99%

Neural dynamics with stochasticity

Siegelmann

1998

Adaptive Processing of Sequences and Data Structures

View full text Add to dashboard Cite

Computational capabilities of recurrent NARX neural networks

Cited by 379 publications

References 24 publications

Estimating Time Series Soil Moisture by Applying Recurrent Nonlinear Autoregressive Neural Networks to Passive Microwave Data over the Heihe River Basin, China

Estimating Time Series Soil Moisture by Applying Recurrent Nonlinear Autoregressive Neural Networks to Passive Microwave Data over the Heihe River Basin, China

Hybrid Neural Network Predictive-Wavelet Image Compression System

Neural dynamics with stochasticity

Contact Info

Product

Resources

About