Corpus ID: 235658388

Reducing numerical precision preserves classification accuracy in Mondrian Forests

  title={Reducing numerical precision preserves classification accuracy in Mondrian Forests},
  author={Marc Vicuna and Martin Khannouz and Gregory Kiar and Yohan Chatelain and Tristan Glatard},
Mondrian Forests are a powerful data stream classification method, but their large memory footprint makes them ill-suited for low-resource platforms such as connected objects. We explored using reduced-precision floating-point representations to lower memory consumption and evaluated its effect on classification performance. We applied the Mondrian Forest implementation provided by OrpailleCC, a C++ collection of data stream algorithms, to two canonical datasets in human activity recognition… Expand

Figures and Tables from this paper


Mondrian Forests: Efficient Online Random Forests
Mondrian forests achieve competitive predictive performance comparable with existing online random forests and periodically retrained batch random forests, while being more than an order of magnitude faster, thus representing a better computation vs accuracy tradeoff. Expand
A Benchmark of Data Stream Classification for Human Activity Recognition on Connected Objects
It is concluded that stream learning for Human Activity Recognition on connected objects is challenged by two factors which could lead to interesting future work: a high memory consumption and low F1 scores overall. Expand
Scalable real-time classification of data streams with concept drift
An empirical study on the serial algorithm of the Micro-Cluster Nearest Neighbour (MC-NN) data stream classifier’s speed, adaptivity and accuracy is provided and an empirical scalability study is provided. Expand
A Study of BFLOAT16 for Deep Learning Training
The results show that deep learning training using BFLOAT16 tensors achieves the same state-of-the-art (SOTA) results across domains as FP32 tensors in the same number of iterations and with no changes to hyper-parameters. Expand
Machine learning method for energy reduction by utilizing dynamic mixed precision on GPU‐based supercomputers
  • K. Rojek
  • Computer Science
  • Concurr. Comput. Pract. Exp.
  • 2019
The achieved results show that the proposed machine learning method allows the accuracy of computation comparable with that achieved double precision and reduce the energy consumption up to 36% compared to the double precision version of MPDATA. Expand
Training Deep Neural Networks with 8-bit Floating Point Numbers
This work demonstrates, for the first time, the successful training of deep neural networks using 8-bit floating point numbers while fully maintaining the accuracy on a spectrum of deep learning models and datasets. Expand
Mining high-speed data streams
This paper describes and evaluates VFDT, an anytime system that builds decision trees using constant memory and constant time per example, and applies it to mining the continuous stream of Web access data from the whole University of Washington main campus. Expand
Multidimensional binary search trees used for associative searching
The multidimensional binary search tree (or <italic>k-d tree) as a data structure for storage of information to be retrieved by associative searches is developed and it is shown to be quite efficient in its storage requirements. Expand
Beating Floating Point at its Own Game: Posit Arithmetic
A new data type called a posit is designed as a direct drop-in replacement for IEEE Standard 754 floating-point numbers (floats), and provides compelling advantages over floats, including larger dynamic range, higher accuracy, better closure, bitwise identical results across systems, simpler hardware, and simpler exception handling. Expand
Verificarlo: Checking Floating Point Accuracy through Monte Carlo Arithmetic
Numerical accuracy of floating point computation is a well studied topic which has not made its way to the end-user in scientific computing. Yet, it has become a critical issue with the recentExpand