[OPINION]:Optimizing chaining function with forward computation transformation

[harisankarsadasivan]

Hello Dr. Li @lh3 ,

Do you think optimizing the chaining function mm_chain_dp with the forward computation transformation from https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8735515 will maintain bit-exact results given you calculate st dynamically and Guo et al cap it at 65? I am observing different results and I personally do not think Guo et al suggestions can maintain accuracy. Would love to hear your thoughts and why you choose to maintain finding the best scoring predecessor in backward direction for chaining.