Developer Guide for Intel® Data Analytics Acceleration Library 2019 Update 5
This Developer Guide documents Intel® Data Analytics Acceleration Library (Intel® DAAL) 2019 Update 5.
The document has been updated to reflect new functionality and enhancements to the product:
DBSCAN, LASSO and Coordinate Descent algorithms were added.
A training alternative was introduced for Logistic and Linear Regressions, SVM, Multi-class Classifier, Decision Forest Classification, GBT Regression and GBT Classification.
New parameters were added to Objective Function interfaces, Sum of Functions and MSE.
Apache Arrow immutable table was added to Numeric Tables.
Changes that have been introduced in the previous update of the document:
New parameter doScale was added to Z-score normalization algorithm.
nClasses parameter was added to BrownBoost, Support Vector Machine, k-Nearest Neighbors, Decision tree, Decision forest, Gradient boosted trees, and Logistic regression classifiers.
Batch processing for logistic regression was updated.
Updates have been introduced to Optimization Solvers, including Iterative Solvers and Objective Function.
Stochastic Average Gradient Accelerated (SAGA) method has been added.
The following objective functions have been updated: MSE, Cross-Entropy Loss and Logistic Loss.
The following iterative solvers have been updated: Stochastic Gradient Descent Algorithm, Limited-Memory Broyden-Fletcher-Goldfarb-Shanno Algorithm and Adaptive Subgradient Method.
New parameter nTrials was introduced for K-means++ initialization.