Recently, I do some literature research about CNN and find there is a concept of

end to end training

Such as the abstract in Fully Convolutional Networks for Semantic Segmentation

How to understand that? What is end-to-end training?

  • $\begingroup$ I understand now. It means that during training, the algorithm does not have separate procedures, just have one step. $\endgroup$ – karl_TUM Sep 4 '16 at 19:18
  • $\begingroup$ +1 for linking to a great paper :-) $\endgroup$ – DeltaIV Jul 5 '18 at 9:33

Traditionally, we extract pre-defined features before prediction.

data -> extracted features -> learning algorithm -> output

However, hand-engineered features limit the potential performance as some of them are poor approximation of reality and some of them throw away some information.

End-to-end learning means that we replace the pipeline with a single learning algorithm so that it goes directly from the input to the desired output to overcome limitations of the traditional approach.

data -> learning algorithm -> output

End-to-end learning system tend to do better when there is a lot of labeled data as the learning algorithm can somehow learn features by itself. When the training set is small, it tends to do worse than hand-engineered pipeline.


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.