a problem with the seed in main_RandomSearch.py

[AdenilsonJr23]

In the code below, the seed is the same for each itr in range(OUT_ITERATION) and each r_itr in range(RS_ITERATION).

tmp_max = get_main.get_valid_performance(DATA, MASK, new_parser, itr, EVAL_TIMES, MAX_VALUE=max_valid)

So the split into trainining, validation and test sets is always the same. There is no cross-validation.
I suggest replacing "seed=1234" with "seed=1234+ itr + 10*r_itr"