Next-Generation Machine Learning Platforms on Many-Core Architectures

Technology Overview

The recent advancement of machine learning technologies (such as deep learning) is not only because of new algorithms that improve accuracy, but also because of new algorithms that exploit the high-performance hardware (such as Graphics Processing Units) and multi-core CPUs to improve efficiency. Our technology aims to help practitioners apply machine learning techniques easily and solve problems at hand quickly. Our technology exploits high-performance hardware (such as GPUs, multi-core CPUs and many-core CPUs) to accelerate the training and inference. Currently, our technology includes a fast Support Vector Machine (SVM) library and another for Gradient Boosting Decision Tree (GBDT) and and Random Forests, while we strive to support more machine learning algorithms.

Technology Features & Specifications

The key features of our technology are listed below. We aim to increase the functionalities to meet the new user demands.

  • Fast: exploit high-performance hardware.
  • Easy-to-use: identical user interface and input options to allow practitioners to easily switch to our technology.
  • Various interfaces: Python, R, MATLAB, etc.
  • Various functionalities: classification, regression, and distribution estimation.
  • Cheaper: achieve higher cost-effectiveness and lower price-performance ratio.

Potential Applications

  • Document classification; email spam filtering
  • Stock price forecasting; time series analysis
  • Network attack detection; anomaly detection
  • Product recommendation; ranking

Customer Benefits

  • High-performance at a low-price
  • Completely transparent to the programmer
  • Switch to exploit high-performance hardware at minimum effort

Make an Enquiry