Tree-based machine learning performed in-memory with memristive analog CAM

Nature Communications, vol. 12, pp. 5806

Abstract

Abstract Tree-based machine learning techniques, such as Decision Trees and Random Forests, are top performers in several domains as they do well with limited training datasets and offer improved interpretability compared to Deep Neural Networks (DNN). However, these models are difficult to optimize for fast inference at scale without accuracy loss in von Neumann architectures due to non-uniform memory access patterns. Recently, we proposed a novel analog content addressable memory (CAM) based on emerging memristor devices for fast look-up table operations. Here, we propose for the first time to use the analog CAM as an in-memory computational primitive to accelerate tree-based model inference. We demonstrate an efficient mapping algorithm leveraging the new analog CAM capabilities such that each root to leaf path of a Decision Tree is programmed into a row. This new in-memory compute concept for enables few-cycle model inference, dramatically increasing 103 × the throughput over conventional approaches.

Authors 8

1. Giacomo Pedretti corresponding

   Hewlett Packard Enterprise (United States)

   Affiliation as printed

   Hewlett Packard Labs, Hewlett Packard Enterprise, Milpitas, CA, USA. giacomo.pedretti@hpe.com

2. Catherine E. Graves corresponding

   Hewlett Packard Enterprise (United States)

   Affiliation as printed

   Hewlett Packard Labs, Hewlett Packard Enterprise, Milpitas, CA, USA. catherine.graves@hpe.com

3. Sergey Serebryakov

   Hewlett Packard Enterprise (United States)

   Affiliation as printed

   Hewlett Packard Labs, Hewlett Packard Enterprise, Milpitas, CA, USA

4. Ruibin Mao

   University of Hong Kong

   Affiliation as printed

   The University of Hong Kong, Hong Kong SAR, China

5. Xia Sheng

   Hewlett Packard Enterprise (United States)

   Affiliation as printed

   Hewlett Packard Labs, Hewlett Packard Enterprise, Milpitas, CA, USA

6. Martin Foltín

   Hewlett Packard Enterprise (United States)

   Affiliation as printed

   Hewlett Packard Labs, Hewlett Packard Enterprise, Milpitas, CA, USA

7. Can Li

   Hewlett Packard Enterprise (United States) · University of Hong Kong

   Affiliation as printed

   Hewlett Packard Labs, Hewlett Packard Enterprise, Milpitas, CA, USA

   The University of Hong Kong, Hong Kong SAR, China

8. John Paul Strachan corresponding Aachen

   Forschungszentrum Jülich · RWTH Aachen University

   Affiliation as printed

   Peter Grünberg Institute (PGI-14), Forschungszentrum Jülich GmbH, Jülich, Germany. j.strachan@fz-juelich.de

   RWTH Aachen University, Aachen, Germany. j.strachan@fz-juelich.de