1
$\begingroup$

A neuroevolution algorithm, such as DXNN, can be used to refine the topology and weights of an artificial neural network (ANN). The GA will require a fitness function, which means you need labeled data for comparison.

Can neuro-evolution be used with unlabelled data? I have no labeled datasets.

|improve this question
$\endgroup$
  • $\begingroup$ Could you give a link or an explanation for feed-forward neuro-modulation? $\endgroup$ – BlindKungFuMaster Apr 13 '17 at 9:13
  • $\begingroup$ @BlindKungFuMaster github.com/CorticalComputer/DXNN2 it is a genetic algorithm that tunes the NN weights and add remove neurons. GA requires to know how each individual performed using their fitness which is established by testing the individual NN solution against a known answer. $\endgroup$ – Aus Apr 13 '17 at 9:14
2
$\begingroup$

The GA will require a fitness function, which means you need labeled data for comparison.

That conclusion is wrong. Yes, sometimes your fitness function will use labeled data. For example, if you want to train an XOR gate or any other known function.

However, there is arguably no advantage of training a function with neuroevolution versus backpropagation, except for the fact that you might discover some new architectures which solve the solution very well.

You don't always need a labelled dataset for neuroevolution.

Take for example IAMDinosaur, which trains neural networks through a genetic algorithm - however, the optimal solution is not known. There is no labelling of input data, all it does is calculate the fitness from the score.

|improve this answer
$\endgroup$
  • $\begingroup$ Got it. That gives me three types of problems, 1- labeled data, 2- unlabeled data but a simulator that gives feedback, 3- no label no feedback. What to do with the a problem of the latter type? like here is a log file, detect what is good and what is bad. Would a classification like SOM works for that? I it feasible? $\endgroup$ – Aus Apr 13 '17 at 10:39
  • $\begingroup$ The neural network in IAMDinosaur maps from sensors to actions (if I understand correctly) and its weights are modified to maximize the rewards it receives from its actions. This makes it a reinforcement learning problem (see Wikipedia for a concise definition). $\endgroup$ – Raketenolli Aug 15 '18 at 20:45

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.