1,SOA pattern effect mitigation by neural network based pre-equalizer for 50G PON
107 48 48 1 (regression) PAM4, learning rate: 0.0001, training symbols:5000, test symbols:5000
2. Machine Learning for 100 Gb/s/λ Passive Optical Network
51 128 128 8 (classification) PAM8, learning rate: -, training symbols:100000, test symbols:90000 cross validation:10000, dropout rate:0.2
3. Training data generation and validation for a neural network-based equalizer
simulation:201 128 128 2 NRZ
experiment:201 128 128 4 PAM4 training symbols:100000, test symbols:100000
4.