High-Performance Data Science


Deliver faster machine learning and analytics results with this library.

Choose & Download

Overview

This library helps reduce the time it takes to develop high-performance data science applications. Enable applications to make better predictions faster and analyze larger data sets with available compute resources.

  • Includes highly optimized machine learning and analytics functions
  • Simultaneously ingests data and computes results for highest throughput performance
  • Supports batch, streaming, and distribution use models to meet a range of application needs
  • Use the same API for application development on multiple operating systems

Note This library supports Python*. To access a Python interface for the Intel® Data Analytics Acceleration Library (Intel® DAAL) high-speed algorithms, use the daal4py that is included in the Intel® Distribution for Python*.

What’s New in the 2020 Edition

  • Introduced new functionality:
    • Probabilistic classification and variable importance computation for gradient boosted trees
    • Classification stump with information gain and Gini index split methods
    • Regression stump with the Mean Squared Error (MSE) algorithm split method
  • Extended existing functionality:
    • Decision tree functionality supports weighted data.
    • AdaBoost algorithm works with multiple classes and algorithms that support weights.
    • AdaBoost multiclass algorithm is available with the SAMME and SAMME.R methods.
  • Improved performance for the L-BFGS Optimization Solver.
  • Deprecation started for new features and functionalities in neural networks. In the 2021 version, support will be completely discontinued. For more information, see the Deprecation Notes.

Release Notes

Fits in the data analytics ecosystem

This library addresses all stages of the data analytics pipeline: preprocessing, transformation, analysis, modeling, validation, and decision-making.

benchmarks for Intel Data Analytics Acceleration Library

About the Library

Intel DAAL outperforms other solutions for developers and data scientists. This benchmark compares performance of the XGBoost implementation in Intel DAAL to an XGBoost open source project. The y-axis shows a speedup factor of two to twelve times for four representative classification and regression test cases.

Specifications

Processors
  • Intel Atom® processors
  • Intel® Core™ processors
  • Intel® Xeon® processors
  • Intel® Xeon Phi™ processors
Code
  • C++
  • Java*
  • Python*
Integrated Development Environments
  • Microsoft Visual Studio* (Windows)
  • Eclipse CDT (C/C++ Development Tooling)*
Operating Systems
  • Windows
  • Linux
  • macOS