weights are needed for targets since often data is unbalance

[Sandy4321]

looks great thanks
only
1
weights are needed for targets since often data is unbalance
2
test should be for very sparse data , since FM is good for sparse data
pls see
https://securityintelligence.com/factorization-machines-a-new-way-of-looking-at-machine-learning/
Factorization machines perform extremely well on sparse data, including data of very high sparsity. As a trade-off, they do not perform well on dense data, so other algorithms are more suited to this class of data.
3
may you add ADAM
they say ADAM is the best
Thanks