I am an experienced programmer but very new to machine learning.

I have a data set that consists of about 50,000 sets of 2,000 ordered values. All of the values are floats normalised to between 0 and 1. The sets of data are analagous to DNA sequences, where there are common patterns between them but each set is a self-contained, discrete set of data.

I am looking for a NN that will be able to predict the remaining 1,900 values in a set when given the first 100.

It appears that an LSTM RNN will be the best model to learn the data - am I correct in thinking this? If I am correct, what would be the best way to process the data? (inputs/layers/nodes/outputs)
The LSTM models I have seen (such as text prediction) seem a little too open-ended for discrete sets of data such as this.

Apologies if this question is a little broad but info on NN on the net is still a little sparse and I need a jumping off point.


I am unable to add a comment so I’ll just write my take; you have sequential data and are trying to infer the remaining sequence given a preceding part of the sequence. This problem can be modeled by a Bayesian Network; check out pomegranate if you are using python, I believe it’s one of the few packages that lets you build easy probability models. For example if you have a sequential data like [[A,B,C,D],[A,B,D,F], etc]](in your case shaped 50000 sequences of length 2000) a Bayesian Network can infer missing values like so [[A,B,None, None]]. The network will infer the missing values. I hope this is what you were looking for.

| cite | improve this answer | |
  • $\begingroup$ hmmm....not heard of that one - I will have to look into it. Thanks for your input $\endgroup$ – chips Sep 13 '19 at 9:15

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.