Machine Learning (ML) is used to exploit some kind of predictability in the system, while in realtime systems you have to know the exact values and their isn't any place for probabilities. So, do you think that there is any kind of ML techniques which could be used in realtime applications?
Hi Bardia,
well I think you need to clarify your question a little bit.
What do you want the ML techniques to do in realtime? Should the learning be done in realtime? Or should the prediction on the basis of a pre-trained model be done in realtime?
Although I am not experienced with realtime applications, there are online algorithms available which can be used to train a model incrementally. Whether or not they are fast enough depends of course on the requirements posed by the environment. The same holds of course for the application phase of the model.
Not all ML techniques deliver probabilities only, there are approaches which deliver numerical or categorical values which are exact, but with different degrees of confidence expressed in probabilities.
Maybe reinforcement approaches are the right thing for your application field, since they will adjust their behaviour to the environment's requirements.
Hope this clarifies things a bit.
Dear Thomas
Thank you
Consider that I have a latency constraint for executing the tasks and I want to predict the future execution‘s makespan.
Hi Bardia Safaei ,
You can find many real-time applications which based-on ML. Robotic systems, image processing applications (pattern recognition, face detection, object detection, etc.), medical industry, etc.
After you clarify your problem, you can search for the ML technique to solve it.
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking