High-SNR Training
A training strategy targeting the high signal-to-noise ratio regime. Through curriculum-based or weighted sampling, it enables a neural decoder to obtain sufficient effective gradients in low error rate regions, thereby maintaining fine-grained decoding performance under clean channel conditions.
Timeline (last 90 days)
The paper proposes two training strategies, progressive training and high-SNR training, which do not change the network structure and only adjust the training procedure.
In the high-SNR regime the bit error rate is extremely low and error samples are scarce, so direct training easily leads to underfitting; the high-SNR training strategy uses curriculum-style or weighted sampling to give the model sufficiently effective gradients in that regime.
All Facts (2)
The paper proposes two training strategies, progressive training and high-SNR training, which do not change the network structure and only adjust the training procedure.
50%UnverifiedIn the high-SNR regime the bit error rate is extremely low and error samples are scarce, so direct training easily leads to underfitting; the high-SNR training strategy uses curriculum-style or weighted sampling to give the model sufficiently effective gradients in that regime.
50%