Histogram algorithm: Discretizes continuous features into about 255 bins. Split finding scans only bins, which is much faster and uses far less memory.
GOSS: Keeps large-gradient samples, randomly samples small-gradient samples and reweights them. Informative samples are kept; long-tail samples are downsampled for speed.
EFB: Bundles mutually exclusive features. High-dimensional sparse features (like one-hot) get merged into one feature to reduce dimension.
Leaf-wise growth: Always splits the leaf with the largest gain. With the same number of leaves it fits stronger, but overfits more easily. On small data, control num_leaves and min_data_in_leaf.
Get asked follow-ups live, then receive a scored report — like a real MLE interview loop.
Start AI mock interview