Back to the Index

Real-Time Machine Learning

What is real-time machine learning?

Real-time ML reference to ML systems where decisions or predictions must be produced with minimal, predictable latency. Real-time ML systems may need very fresh features, which can be on-demand features or precomputed features created by a streaming feature pipeline. Batch features may be used if the model is not sensitive to stale feature values. Real-time ML systems have either:

  • online inference pipelines, where an interactive application expects a prediction response to its request within some bounded time period, e.g., a 50ms SLA, or 
  • streaming inference pipelines, where a non-interactive system sends prediction requests to an event bus and gets responses created by a streaming inference pipeline, e.g., running in Apache Flink or Spark Streaming.
Does this content look outdated? If you are interested in helping us maintain this, feel free to contact us.

© Hopsworks 2024. All rights reserved. Various trademarks held by their respective owners.

Privacy Policy
Cookie Policy
Terms and Conditions