can anyone please give me some idea or links as to how the autoregressive neural networks work for time series?

I tried to go through the thesis of Sebastian Dietz, but I could not able to understand the network intuitively.

Any help is appreciated, thanks in advance.


I suggest this video of Hugo Larochelle.

In a nutshell, autoregressive neural network are designed to predict some dimensions given a subset of other dimensions; for example in images, to predict one pixel given a subset of other pixels; for time series, to predict one time sample given a subset of time samples.

Note that for time series, some models do not restrict to the natural ordering of the dimensions, and are trained to predict one time sample given both past and future time samples.

Examples includes NADE, PixelCNN, PixelRNN, ...


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.