We also make several improvements to decoding, employing two types of offline high-accuracy decoders. One is a neural network decoder [27],
I’m very suspicious of this. Does anybody understand what the point of this is? How is a decoder being implemented as a neural network - won’t that simply be a device for “guessing” and be prone to it’s own hallucination?
How can you really trust your error correction “algorithm” if a significant part, the decoding, is done by a black box neural network that may or may not give accurate results?
I’m very suspicious of this. Does anybody understand what the point of this is? How is a decoder being implemented as a neural network - won’t that simply be a device for “guessing” and be prone to it’s own hallucination?
How can you really trust your error correction “algorithm” if a significant part, the decoding, is done by a black box neural network that may or may not give accurate results?
[Comment removed by author]
this doesn’t answer my question positively, it just tries to dismiss my question negatively
[Comment removed by author]