Abstract:We propose a novel end-to-end (E2E) fiber-terahertz (THz) integrated system based on a data-driven mutual information neural estimator (MINE). This system utilizes the MINE neural networks for Tx-NNs optimization, enabling constellation shaping, channel sampling without known channel models, and a differentiable carrier phase estimation for resisting phase noise. Our approach follows state-of-the-art bitwise autoencoders (AE), which requires a differentiable implementation of all operations between transmitter and receiver, including the DSP algorithms. Firstly, to address the demand for higher spectral efficiencies, we employ the E2E-AE framework for constellation shaping in the presence of laser noises and nonlinearity. Subsequently, we rigorously examine the characterization of lasers with white and colored frequency modulation noise and investigate their impact on MI performance using simulation and experiment when the carrier phase recovery algorithm is present and absent. Finally, We experimentally evaluate our proposed E2E framework within a dual-polarized fiber-THz integrated system at 320GHz, considering linewidth, flicker ( $1/f$ ) noise, and resonance peaks. By joint optimization constellation shaping and compensation for the phase noise and nonlinearity, we achieve a robust and pilot-free modulation scheme improving the performances by up to 0.15-bit and 1.21-dB gains in MI and SNR compared to Grid-QAM constellations. Additionally, under the constraints of laser phase noise and nonlinearity, we achieve a transmission speed of 230Gbit/s in the E2E-AE fiber-THz system for the first time. This solution advances the E2E-AE framework by incorporating differentiable functions tailored for practical scenarios.

Autoencoder Learning of Constellation Shaping Robust to Semiconductor Laser Noise and Nonlinearity in Fiber-THz System

Autoencoder Learning of Nonlinear Constellation Shape for Fiber-Wireless Convergence System.

End-to-End Optimization of Constellation Shaping Robust to Semiconductor Laser Noise and Nonlinearity in Fiber-THz System

End-to-End Learning Based Symbol-to-Symbol Autoencoder for G-band Fiber-Terahertz Integrated Communication System

Nonlinearity Mitigation in a Fiber-Wireless Integrated System Based on Low-Complexity Autoencoder and BiLSTM-ANN Equalizer.

End-to-end Learning of a Constellation Shape Robust to Channel Condition Uncertainties

Demonstration of photonic neural network for fiber nonlinearity compensation in long-haul transmission systems

An Autoencoder-Based Constellation Design for AirComp in Wireless Federated Learning

Overcoming laser phase noise for low-cost coherent optical communication

Auto-Encoder Optimized PAM IM/DD Transceivers for Amplified Fiber Links

Autoencoder with Fitting Network for Terahertz Wireless Communications: A Deep Learning Approach

Autoencoder for Optical Wireless Communication System in Atmospheric Turbulence

Model-Based Deep Learning of Joint Probabilistic and Geometric Shaping for Optical Communication

End-to-End Optimization and Equalization Based on Deep-Learning for Fiber-Terahertz Integrated Communication System at 209 GHz

Deep Learning-Based End-to-End Bit-Wise Autoencoder for G-Band Fiber-Terahertz Integrated DFT-S-OFDM Communication System.

Adaptive feedback-driven probabilistic shaping for high-order modulation formats in a fiber-THz seamless integration system at 320 GHz

Robust Semantic Communications Against Semantic Noise

Improving decryption quality of optical chaos communication using neural networks

Model-Free End-to-End Deep Learning of Joint Geometric and Probabilistic Shaping for Optical Fiber Communication in IM/DD System

An Optimal Adaptive Constellation Design Utilizing an Autoencoder-Based Geometric Shaping Model Framework

Two-dimensional End-to-end Deep Learning Autoencoder in G-Band Fiber-Terahertz Integrated Transmission for 6G RAN