Say I trained a Neural Network (not RNN or CNN) to classify a particular data set.

So I train using a specific data set & then I test using another and get an accuracy of 95% which is good enough.

I then deploy this model in a production level environment where it will then be processing real world data.

My question is, will this trained NN be constantly learning even in a production scenario? I can't figure out how it will because say it processes a dataset such as this: [ [1,2,3] ] and gets an output of [ 0, 0.999, 0 ]

In a training scenario it will compare the predicted output to the actual output and back propagate but in a real world scenario it will not know the actual value.

So how does a trained model learn in a real world scenario?

I am still very much a beginner in this field and I am not sure if the technology used is going to affect the answer to this question, but I am hoping to use Eclipse Deeplearning4J to create a NN. That being said the answer does not need to be restricted to this technology in particular as I am hoping more for the theory behind it and how it works.


You are right. If you don't continuously train the neural network after you have deployed it, there is no way it can continuously learn or be updated with more information. You need to program the neural network to learn even after it has been deployed. There is no such thing as a neural network that decides what it does without a human deciding first what it needs to do: this is a very common misconception (probably caused by the media and science fiction movies). It's also true that you need to label your data if you intend to train the neural network in a supervised fashion, but there are other ways to train neural networks (e.g. by reinforcement learning), depending also on the problem you want to solve.

If you want to develop neural networks that can learn continually, you probably want to look into continual learning techniques for neural networks. Another term that you may be looking for is online machine learning.

  • $\begingroup$ Thanks for the answer and the edit! "incremental learning" and "continuous learning" learning were the terms I was looking for. Saved the linked research paper for a later day as it looked slightly intimidating at present. $\endgroup$ – user3536523 Sep 26 '20 at 12:12
  • $\begingroup$ @user3536523 No problem. Yes, I suggest that you first get familiar with the basic concepts of machine learning and neural networks before reading that paper, although it's not a very difficult paper to read, but you would not understand many concepts if you are not familiar with the basics. $\endgroup$ – nbro Sep 26 '20 at 12:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.