A ml beginner here, so please bear with me. If I understand correctly RNNs seem to be the go to method right now for sequence prediction for a given input (single/as a sequence). But I do not have sufficient data to train a RNN. I have discounted Markov decision process based mechanisms for the same reason.

Are there any online learning algos that I can use to get coarse/approximate predicted sequences with only a small training set? I have looked at Q-learning but it seems to be ideal for best path problems where the end goal is definite.

Any pointers would be greatly appreciated.

UPDATE: Adding more clarity on the type of data post the comments. My data is video content consumption data. About 100 users consuming from a library of 1000 video titles. Intent is to exploit (if it exists) the likelihood of consuming content in a probable sequence. I currently have 300 - 400 such sequences spanning 3 videos each.

  • $\begingroup$ Welcome to DataScience.SE! The traditional way to deal with limited data is to introduce assumptions through Bayesian priors, or stronger regularization. What do you know about your data generation process? $\endgroup$
    – Emre
    Jun 20 '16 at 18:36
  • $\begingroup$ Could you tell me how small is your training set? Basically if you could provide more details on regarding your problem would be nice. $\endgroup$
    – ahajib
    Jun 20 '16 at 20:22

Your Answer

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

Browse other questions tagged or ask your own question.