Temporal walks vs Ngram

[sameeravithana]

Can I feed unique temporal walks (i.e., random walks forward in the time) as ngrams to calculate the optimal k-th order of the underlying network?

[verginer]

Hi @SamTube405 thanks for reaching out!
I am not sure I understand your question correctly.
You have data only on the paths taken, whatever they may be, but not how often, but at least once?
If you have a strong theoretical or empirical argument that the system under consideration is well represented by these paths then maybe. However, most likely the underlying system is not sufficiently characterised by its unique paths and hence you would not estimate what you want.
In other words the data is not sufficient to identify memory effects.