Abstract
This chapter describes a variety of techniques for writing efficient, scalable, and general-purpose decision forest software. It will cover:
-
algorithmic considerations, such as how to train in depth first or breadth first order;
-
optimizations, such as cheaply evaluating multiple thresholds for a given feature;
-
designing for multi-core, GPU, and distributed computing environments; and
-
various ‘tricks of the trade’, including tuning parameters and dealing with unbalanced training sets.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The frontier can comprise all the leaf nodes or only a subset. It might contain leaf nodes from different levels in the tree. Some leaf nodes might not be on the frontier; for example, those nodes for which no candidate weak learner could improve the information gain at a previous iteration of breadth first training.
- 2.
Of course, for rigorous testing, you should always use a hold-out validation set when optimizing these parameters: optimizing on the test set will lead to misleadingly inflated accuracy scores.
- 3.
A similar strict separation is even more important to ensure truly independent training and test sets.
References
Breiman L (1996) Bagging predictors. Mach Learn 24(2)
Budiu M, Shotton J, Murray D, Finocchio M (2011) Parallelizing the training of the Kinect body parts labeling algorithm. In: Advances in neural information processing systems (NIPS). BigLearn workshop
Geurts P, Ernst D, Wehenkel L (2006) Extremely randomized trees. Mach Learn 36(1)
Oberg J, Eguro K, Bittner R, Forin A (2012) Random decision tree body part recognition using FPGAs. In: Proc 22nd int conf on field programmable logic and applications (FPL)
Sharp T (2008) Implementing decision trees and forests on a GPU. In: Proc European conf on computer vision (ECCV). Springer, Berlin
Shotton J, Girshick R, Fitzgibbon A, Sharp T, Cook M, Finocchio M, Moore R, Kohli P, Criminisi A, Kipman A, Blake A (2012) Efficient human pose estimation from single depth images. IEEE Trans Pattern Anal Mach Intell
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2013 Springer-Verlag London
About this chapter
Cite this chapter
Shotton, J., Robertson, D., Sharp, T. (2013). Efficient Implementation of Decision Forests. In: Criminisi, A., Shotton, J. (eds) Decision Forests for Computer Vision and Medical Image Analysis. Advances in Computer Vision and Pattern Recognition. Springer, London. https://doi.org/10.1007/978-1-4471-4929-3_21
Download citation
DOI: https://doi.org/10.1007/978-1-4471-4929-3_21
Publisher Name: Springer, London
Print ISBN: 978-1-4471-4928-6
Online ISBN: 978-1-4471-4929-3
eBook Packages: Computer ScienceComputer Science (R0)