Neural network approaches to capture temporal information

Veelen, Martijn van; Nijhuis, J.A.G.; Spaanenburg, Ben

doi:10.1063/1.1291274

Cited by 8 publications

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The output of each tap in the delay line is determined as [8] The memory depth of an ordinary TDL is fixed and equal to . In contrast, since the memory resolution is an adaptive parameter, the memory depth of a Gamma TDL can be adapted to minimize the output mean square error using Gradient descent procedure during the training process.…”

Section: Proposed Fully Rnn-based Modelmentioning

confidence: 99%

A Fully Recurrent Neural Network-Based Model for Predicting Spectral Regrowth of 3G Handset Power Amplifiers With Memory Effects

Luongvinh

Kwon

2006

IEEE Microw. Wireless Compon. Lett.

View full text Add to dashboard Cite

Efficient and accurate behavioral models of power amplifiers (PAs) with memory effects are important for predicting the distortions generated by PAs in 3G handsets. Conventional recurrent neural network (RNN) has been applied for RF PAs, but its capability to model PAs with memory effects has not been investigated. In this letter, we propose a new fully RNN with Gamma tapped-delay lines suitable for modeling the dynamic behavior of 3G PAs with memory effects. After being trained with wideband code division multiple access (W-CDMA) (3GPP Uplink) signals, the proposed model is validated with not only W-CDMA but also high-speed downlink packet access (3GPP Uplink) signals with higher peak-to-average ratios (PARs), which demonstrates the generality of the model. The comparisons with previous RNN models show that the proposed model offers improved performance in predicting spectral regrowth by reducing errors by 1.7-4 dB.Index Terms-Adjacent channel leakage ratio (ACLR), behavioral modeling, high-speed downlink packet access (HSDPA), memory effects, recurrent neural network (RNN).

show abstract

Section: Proposed Fully Rnn-based Modelmentioning

confidence: 99%

A Fully Recurrent Neural Network-Based Model for Predicting Spectral Regrowth of 3G Handset Power Amplifiers With Memory Effects

Luongvinh

Kwon

2006

IEEE Microw. Wireless Compon. Lett.

View full text Add to dashboard Cite

show abstract

“…A further issue to be addressed here is the inclusion of time (serial correlation) in NNs. Van Veelen et al (2000) review the different solutions applicable to NNs dealing with time series data. The authors present two main approaches to the inclusion of time information in NNs:…”

Section: Neural Forecastingmentioning

confidence: 99%

“…The alternative approach of explicitly including time information is also criticized (see Van Veelen et al, 2000), since it does not include a dynamic framework. This shortcoming may possibly result in the incapacity of NNs to locate hidden time trends.…”

Section: Neural Forecastingmentioning

confidence: 99%

Neural Networks and Genetic Algorithms as Forecasting Tools: A Case Study on German Regions

Patuelli

Longhi

Reggiani

et al. 2008

Environ Plann B Plann Des

View full text Add to dashboard Cite

This paper develops and applies neural network (NN) models to forecast regional employment patterns in Germany. Computer-aided optimization tools that imitate natural biological evolution to find the solution that best fits the given case (namely, genetic algorithms, GAs) are also used to detect the best NN structure. GA techniques are compared with more ‘traditional’ techniques which require the supervision of experienced analysts. We test the performance of these techniques on a panel of 439 districts in West and East Germany. Since the West and East datasets have different time spans, the models are estimated separately for West and East Germany. The results show that the West and East NN models perform with different degrees of precision, mainly because of the different time spans of the two datasets. Automatic techniques for the choice of the NN architecture do not seem to outperform selection procedures based on the supervision of expert analysts.

show abstract

“…Since the late 1990s machine learning methods have been used in time series prediction. These methods have demonstrated to be powerful non-linear estimators applicable to many real-world problems [2]. Various Machine learning paradigms have approached the time series prediction problem, e.g.…”

Section: Introductionmentioning

confidence: 99%

Time Series Prediction Using Restricted Boltzmann Machines and Backpropagation

Hrasko

Pacheco

Krohling

2015

Procedia Computer Science

View full text Add to dashboard Cite

Time series prediction appear in many real-world problems, e.g., financial market, signal processing, weather forecasting among others. The underlying models and time series data of those problems are generally complex in a way that reasonable accurate estimation cannot be easily achieved, thus requiring more advanced techniques. Statistical models are the classical approaches for tackling this problem. Many works extended different architectures of Artificial neural networks to work with time series prediction, such as Feedforward, Boltzmann Machines and Deep Belief Network. A Deep Belief Network based on hybridization between Gaussian-Bernoulli Restricted Boltzmann Machine and the Backpropagation algorithm is presented. The hybrid algorithm is tested on three time series databases showing promising results.

show abstract

Neural network approaches to capture temporal information

Cited by 8 publications

References 21 publications

A Fully Recurrent Neural Network-Based Model for Predicting Spectral Regrowth of 3G Handset Power Amplifiers With Memory Effects

A Fully Recurrent Neural Network-Based Model for Predicting Spectral Regrowth of 3G Handset Power Amplifiers With Memory Effects

Neural Networks and Genetic Algorithms as Forecasting Tools: A Case Study on German Regions

Time Series Prediction Using Restricted Boltzmann Machines and Backpropagation

Contact Info

Product

Resources

About