New Adaptive Filter Improves Time-Series Prediction with Input Noise

Yi Peng, Haiquan Zhao, Jinhui Hu· July 23, 2026 View original

Summary

Researchers developed the RFFBCGA algorithm, a new nonlinear adaptive filter that effectively mitigates both input and output noise in time-series prediction. This method maintains a fixed network structure while enhancing robustness across various noise scenarios.

Traditional nonlinear adaptive filtering often overlooks input noise, focusing primarily on output noise. While the existing BCKLMS algorithm addressed input noise, it struggled with fixed dictionary sizes limiting signal capture and poor robustness against non-Gaussian output noise. A new algorithm, RFFBCGA (Random Fourier Bias-Compensated Filter under General Adaptive Function), has been proposed to overcome these limitations. It uses a random Fourier feature-based framework to maintain a fixed network structure and effectively reduce input noise. Furthermore, by incorporating a flexible general adaptive function, RFFBCGA significantly improves robustness against diverse noise types in the output signal. Extensive simulations, including real-world time series prediction, demonstrate its superior performance.

Why it matters

Professionals working with noisy time-series data in fields like finance, engineering, or environmental monitoring can achieve more accurate predictions and robust model performance.

How to implement this in your domain

  1. 1Investigate integrating RFFBCGA into existing time-series forecasting pipelines for improved accuracy.
  2. 2Evaluate the algorithm's performance against current models using datasets with known input and non-Gaussian output noise.
  3. 3Collaborate with research teams to adapt and optimize the RFFBCGA framework for specific industry applications.
  4. 4Develop internal expertise in advanced adaptive filtering techniques to leverage such innovations.

Who benefits

FinanceManufacturingEnergyEnvironmental MonitoringHealthcare

Key takeaways

  • The RFFBCGA algorithm offers enhanced time-series prediction by addressing both input and output noise.
  • It improves robustness against non-Gaussian noise, a common challenge in real-world data.
  • The method maintains a fixed network structure while effectively capturing input signal characteristics.
  • This research provides a superior alternative to previous bias-compensated adaptive filtering techniques.

Original post by Yi Peng, Haiquan Zhao, Jinhui Hu

"arXiv:2607.19902v1 Announce Type: new Abstract: Most existing nonlinear adaptive filtering algorithms only account for output noise, neglecting the fact that input noise is also prevalent in practice. Although the recently proposed bias-compensated kernel least mean square (BCKLM…"

View on X

Originally posted by Yi Peng, Haiquan Zhao, Jinhui Hu on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI ResearchAI Engineering & DevTools

New Algorithm Learns Local Causal Structures with Latent Variables

Researchers propose LoCaLS, a new algorithm for learning local causal structures around a target variable from observational data, even when latent variables and selection bias are present. LoCaLS achieves high accuracy with significantly less computational effort than global causal discovery methods.

Zheng Li, Hao Zhang, Ruxin Wang, Ruichu Cai, Kun Zhang, Feng XieJul 23, 2026
AI Engineering & DevToolsAI Research

New Framework Evaluates AI Robustness with Minimum-Norm Attacks

Researchers introduce a unified framework for evaluating adversarial robustness using a comprehensive pool of minimum-norm attacks and robustness-perturbation curves across multiple norms. This approach addresses limitations of fixed-epsilon evaluations, providing a more stable and controllable assessment of AI model defenses.

Luca Scionis, Luca Melis, Maura Pintor, Fabio Brau, Ambra Demontis, Giorgio Fumera, Fabio Roli, Battista BiggioJul 23, 2026
AI ResearchAI Engineering & DevTools

New RL Algorithm Achieves Horizon-Free Optimal Regret

Researchers propose a new reinforcement learning algorithm for finite-horizon time-homogeneous tabular Markov Decision Processes that achieves asymptotically optimal regret without dependence on the horizon length (H). This significantly improves upon previous guarantees by removing the log H factor and drastically improving the asymptotic bound.

Runlong Zhou, Zihan Zhang, Maryam Fazel, Simon S. DuJul 23, 2026