Part 2: Advanced scikit-learn* Essentials for Machine Learning

Overview

This is the final part of a two-part workshop series that demonstrates how to:

  • Speed up key machine learning algorithms that rely on scikit-learn*
  • Get faster results without specialized hardware

This workshop builds on Part 1: Introduction to scikit-learn* Essentials for Machine Learning by extending the capability using the compute follows data method of taking advantage of Intel® Extension for Scikit-learn* for current and upcoming Intel GPUs.

In this more advanced workshop, you have the opportunities to:

  • Practice what you learned in Part 1 to speed up two codes: an image-clustering example and a galaxy-collision example
  • Revisit patching strategies to ensure that the performance of your scikit-learn code runs faster than stock scikit-learn
  • Apply patching strategies with coarse-grained to surgical-level precision
  • Use a special technique to perform scikit-learn computation on Intel GPUs

 

Jump to:

You May Also Like
 

Intel® Extension for Scikit-learn*

Speed up and scale your scikit-learn* workflows for CPUs and GPUs across single- and multi-node configurations with this Python* module for machine learning.

 

 

See All Tools

 

Highlights

0:00 Introductions

1:32 Learning objectives

3:05 Programming challenges

4:04 oneAPI initiative

5:26 Intel® oneAPI toolkits

5:50 Intel oneAPI ecosystem

8:28 Motivation for today

10:35 Computer follows data

12:25 Procedure for ignition

14:04 Gallery of algorithms optimized for Intel CPUs and GPUs

19:16 Patching and imports: The order

20:16 Train and test split: The syntax

21:00 Prepare for computer follows data: dpctl

21:53 Device queue

23:11 Apply a patch to scikit-learn

23:48 Prepare data

26:54 Casting

29:03 When to cast data returned from the device

29:37 Intel® Developer Cloud access

30:22 Intel Developer Cloud: Enroll and sign in

31:36 Intel Developer Cloud: Jupyter* lab access

50:05 Exercise: Module 05_01, introduction to Intel Extension for Scikit-learn

1:12:00 Queue function

1:16:45 scikit-learn K-means

1:36:00 Polling questions

1:37:17 Apply the patch

1:41:40 Exercise: Module_05_02, K-nearest neighbor (KNN): targeting a GPU and patching

1:47:37 Casting a NumPy array and predicting

1:52:40 Exercise: Module 05_03, a gallery of functions on GPUs