Hi guys I have a slight confusion on the learning process of neural networks.

When the input layer receives inputs, goes through the hidden layers and then into the output layer. How does the neural network know that the outputs at the output layer are incorrect?

When the error is calculated at the output layer, it needs the predicted output which is fine but how is the actual output found?



Well, you make a very good point, this is the most painful part of machine learning, we have to label data to make them usable to train models.By labelling the data, we mean to go through the dataset and say what the output of each input is. It means that to each input we attribute a true output.

The input of the learning function in sk-learn for example are always (almost) model.train(X, Y). with X the inputs and Y the corresponding expected outputs.

Labelling is a tedious and long task that must be hand made most of the time. Some website even pay people to label data (especially for image classification or segmentation), such as amazon mechanical turk.


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.