Transient Simulations of High-Speed Channels Using CNN-LSTM With an Adaptive Successive Halving Algorithm for Automated Hyperparameter Optimizations

Goay, Chan Hong; Ahmad, Nur Syazreen; Goh, Patrick

doi:10.1109/access.2021.3112134

Cited by 22 publications

(8 citation statements)

References 36 publications

(32 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [18], the authors introduced an algorithm named adaptive successive halving automated hyperparameter optimization (ASH-HPO), integrating successive halving, Bayesian optimization, and progressive sampling. This method was used to tune hyperparameters for RNN models, specifically for transient simulations of high-speed channels.…”

Section: Bayesian Optimization Methodsmentioning

confidence: 99%

Applying the Deep Learning Techniques to Solve Classification Tasks Using Gene Expression Data

Babichev,

Liakh,

Kalinina

2024

IEEE Access

View full text Add to dashboard Cite

This manuscript explores the application of deep learning (DL) techniques for classifying gene expression data. A key aspect of our research is the comparative analysis of various DL neural network architectures, including Convolution Neural Networks (CNN), Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) Recurrent Neural Networks (RNN), as well as hybrid models that combine these networks. We applied the Bayesian optimization algorithm using 5-fold cross-validation for optimal hyperparameter tuning, which is crucial for DL algorithm performance. Significantly, we have advanced the methods for applying RNNs in processing gene expression data, particularly focusing on LSTM and GRU types. Our study introduces also a novel hybrid quality criterion for data classification, calculated as a weighted sum of partial quality criteria, incorporating an integrated F1-score derived through the Harrington desirability method. Furthermore, we investigate hybrid models that leverage various DL methods, enhancing decision-making objectivity in sample identification. This model uses a step-by-step information processing procedure, initially applying different DL models to gene expression data and subsequently processing these through a CART-based classifier for final decision-making. Our experiments, performed on gene expression data from patients with eight cancer types and one subset with normal samples (without cancer), demonstrated that GRU-RNN-based models, particularly a two-layer GRU-RNN, achieved the highest classification efficacy, with an accuracy of 97.8% on the test dataset. The performance of this model exceeded that of other models, whose accuracy varied between 96.6% and 97.3%. Comparative analysis with other studies in this field suggests that the proposed techniques demonstrate higher efficacy compared to similar research regarding the application of DL models for cancer-type diagnosis.

show abstract

Section: Bayesian Optimization Methodsmentioning

confidence: 99%

Applying the Deep Learning Techniques to Solve Classification Tasks Using Gene Expression Data

Babichev,

Liakh,

Kalinina

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Due to the high-speed movement of the train, HSR channel exhibit a high Doppler frequency shift and a short coherence time. The CSI predicted using the conventional AR sliding window model encounters the problem of inaccurate precoding effects, attributed to the small coefficients of the model [23]. Therefore, the conventional AR sliding window model is not suitable for precoding predictions.…”

Section: B Ar-ps Precoding Processmentioning

confidence: 99%

“…Employing the progressive sampling method conducts sampling the HSR channel at adjacent relatively short intervals. It is essential to note that progressive sampling is not random sampling; rather, it initiates from the first sampling time sequence and proceeds along the time sequence [23].…”

Section: B Ar-ps Precoding Processmentioning

confidence: 99%

“…Nevertheless, LSTM networks tackle the vanishing gradient problem in RNNs networks by incorporating forget gates and peephole connections. [7] proposed a CNN-LSTM structure for channel prediction, utilizing CNN for feature extraction and employing LSTM networks for sequence prediction, to speed up the transient simulations of high-speed channel. In [8], a novel RNNbased MIMO channel predictor was proposed.The predictor combined LSTM network and Gated Recurrent Unit (GRU) to improve channel prediction accuracy.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Time-frequency domain joint precoding based on the LSTM network and repetitive HSR channel

Zhu,

Duan,

Xue

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…This approach however necessitates the utilization of sophisticated tools and hardware, potentially limiting its applicability to smaller or less complex control systems. Previous studies have also explored the use of nonlinear controllers such as neural networks [22]- [25], fuzzy logic [26]- [29], backstepping [30], deep reinforcement learning [31], and passivity-based control [32], as alternative strategies for stabilizing the BnB system.…”

Section: Introductionmentioning

confidence: 99%

Modeling and Hybrid PSO-WOA-Based Intelligent PID and State-Feedback Control for Ball and Beam Systems

Ahmad

2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

The ball and beam (BnB) system serves as a benchmark in control engineering as it provides a foundational concept applicable to addressing stabilization challenges of various underactuated nonlinear systems. This includes tasks like maintaining the balance of goods carried by mobile robots and controlling the attitude of unmanned aerial vehicles. In this study, the focus is on enhancing control optimization strategies for BnB systems that take into account inherent nonlinearities arising from actuator constraints and state measurements. The work introduces a novel intelligent control approach, termed hybrid PSO-WOA, which combines Particle Swarm Optimization (PSO) and Whale Optimization Algorithm (WOA) to automate optimal parameter search for proportional-integral-derivative (PID) and state feedback (SF) controllers. The collaborative technique between PSO and WOA is formulated to strike a balance between exploration and exploitation phases, and to mitigate premature convergence risks due to the system's complexities. Additionally, three control schemes, namely cascade PID-PID, cascade PID-SF, and cascade PID-observer are introduced, each with tailored cost functions for optimization through the hybrid PSO-WOA algorithm, accommodating both measurable and unmeasurable state scenarios. Simulation results consistently demonstrate the superior performance of the hybrid approach compared to individual PSO and WOA methods, as well as conventional PID and linear quadratic regulator approaches. Notable, the hybrid approach exhibits a significant improvement in error metrics, reducing integral-time absolute error by 18.99%, integral squared error by 35.37%, and steady-state error by 92.86%. This substantial enhancement suggests promising directions for future research in automated control parameter tuning for underactuated nonlinear systems.

show abstract

Transient Simulations of High-Speed Channels Using CNN-LSTM With an Adaptive Successive Halving Algorithm for Automated Hyperparameter Optimizations

Cited by 22 publications

References 36 publications

Applying the Deep Learning Techniques to Solve Classification Tasks Using Gene Expression Data

Applying the Deep Learning Techniques to Solve Classification Tasks Using Gene Expression Data

Time-frequency domain joint precoding based on the LSTM network and repetitive HSR channel

Modeling and Hybrid PSO-WOA-Based Intelligent PID and State-Feedback Control for Ball and Beam Systems

Contact Info

Product

Resources

About