1：Experimental Design and Method Selection:

The experiment involves transmitting data over a 10-m SI-POF using M-PAM and employing a multilayer perceptron-based equalizer to mitigate ISI and non-linearity. The conventional transversal DFE is also used for comparison.

2：Sample Selection and Data Sources:

A uniformly distributed random binary data of length 106 bits is used, preceded by a preamble of 4000 symbols for synchronization and training.

3：List of Experimental Equipment and Materials:

- Optical source: Red laser diode (LD, L650P007, 650 nm wavelength, numerical aperture ≈ 0.15, 7 mW optical power).

4：15, 7 mW optical power). - Receiver:

4. Experimental Procedures and Operational Workflow: The binary data is M-PAM modulated, upsampled, and fed through a digital pulse shaping filter. The modulated signal is then loaded to an AWG, transmitted through the SI-POF, received, converted to an electrical signal, captured, and post-processed for BER evaluation.

5：45 A/W. - SI-POF:

5. Data Analysis Methods: The BER is evaluated for different training examples and PAM levels, comparing the performance of MLP-DFE and TR-DFE.