Closed Luke100000 closed 5 days ago
Two alternative losses which may help, parallelism and a LightGBM implementation. They are usually easier to tune, significantly faster to train, work on tabular categorical data directly, and perform quite well.
Two alternative losses which may help, parallelism and a LightGBM implementation. They are usually easier to tune, significantly faster to train, work on tabular categorical data directly, and perform quite well.