What is weighted quantile sketch?
And a weighted quantile sketch merges the data into an approximate histogram, then the histogram divided into weighted quantiles that put low confidence predictions into smaller size quantiles.
Why XGBoost?
XGBoost is a scalable and accurate implementation of gradient boosting machines and it has proven to push the limits of computing power for boosted trees algorithms as it was built and developed for the sole purpose of model performance and computational speed.
Why XGBoost is better than deep learning?
Xgboost is an interpretation-focused method, whereas neural nets based deep learning is an accuracy-focused method. Xgboost is good for tabular data with a small number of variables, whereas neural nets based deep learning is good for images or data with a large number of variables.
What is quantile sketch?
Quantiles sketch is a mergeable streaming algorithm to estimate the distribution of values, and approximately answer queries about the rank of a value, probability mass function of the distribution (PMF) or histogram, cumulative distribution function (CDF), and quantiles (median, min, max, 95th percentile and such).
Why are XGBoost and normal quantiles the same?
In XGBoost, the Weighted Quantile Sketch will be little bit different. The thing is it will use a special WEIGHTED quantile instead of using the normal quantile. For normal quantile, the number of observations are the same of each quantile. In contrast, the weighted quantile might not be the same.
How to understand the weighted quantile sketch algorithm?
To understand the weighted quantile sketch we have to know quantile sketch first, we split the huge dataset into multiple small pieces and put the pieces on different computers on a network. The Quantile Sketch Algorithm combines the values from each computer to make a rough histogram.
When to use parallel learning and XGBoost?
To conclude, approximate greedy algorithm, parallel learning and weighted quantile sketch will only be used when the datasets are extremely large. When the training data is small, XGBoost will just use normal greedy algorithm.
When do you use XGBoost for sharding?
Also, when there are more than one Hard Drive available for storage, XGBoost will use Sharding to accelerate the disk access. To conclude, approximate greedy algorithm, parallel learning and weighted quantile sketch will only be used when the datasets are extremely large.