id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_les4de22qzd4pkbxjnruyovkn4	Rory Mitchell	Accelerating the XGBoost algorithm using GPU computing	2017	37	.pdf	application/pdf	14157	1608	70	interleaved subsets of data on GPUs and develop a massively parallel tree construction XGBoost is at its core a decision tree boosting algorithm. by scanning left to right through all feature values in a leaf in sorted order. XGBoost algorithm handles this by performing two scans over the input data, the second node, update the positions of training instances based on these new splits and then The first phase of the algorithm finds the best split for each leaf node at the current level. GPU algorithm is to perform a sum reduction over the entire feature before scanning. algorithm for a thread block processing a single feature at a given tree level is shown in The sorting implementation of the split finding algorithm operates on feature value data Given data sorted by node ID first and then feature values The performance and accuracy of the GPU tree construction algorithm for XGBoost is	./cache/work_les4de22qzd4pkbxjnruyovkn4.pdf	./txt/work_les4de22qzd4pkbxjnruyovkn4.txt