BLSTM convolution and self-attention network enabled recursive and direct prediction for optical chaos.

Area of Science:

• Optics
• Nonlinear Dynamics
• Artificial Intelligence

Background:

• Chaotic time series prediction is crucial for applications like secure communications and random number generator analysis.
• Existing methods face challenges with error accumulation in long-term predictions.

Purpose of the Study:

• To develop an advanced model for accurate optical chaos prediction.
• To improve the prediction accuracy and extend the prediction duration of chaotic time series.

Main Methods:

• A deep learning model combining Bidirectional Long Short-Term Memory (BLSTM) networks, convolution, and self-attention mechanisms was proposed.
• The model was validated for both direct and recursive prediction scenarios.
• Performance was evaluated using Normalized Mean Squared Error (NMSE).

Main Results:

• The proposed BLSTM convolution and self-attention network model significantly reduced error accumulation.
• The model achieved accurate predictions for optical chaos over extended time durations, reaching 4 ns.
• A low Normalized Mean Squared Error (NMSE) of less than 0.01 was obtained.

Conclusions:

• The developed model offers a powerful tool for precise optical chaos prediction.
• This approach enhances the reliability and scope of chaotic time series analysis.
• The findings have implications for secure communication systems and random number generation.