aws
deepar dynamic feature definition please
it is not clear what dynamic feature is? I can not find description/definition on https://docs.aws.amazon.com/sagemaker/latest/dg/deepar-in-formats.html and in paper not mentioned ? https://arxiv.org/pdf/1704.04110.pdf
if somebody can share link to definition what is it?
can it be categorical value?
we see from page 84 https://gmoein.github.io/files/Amazon%20SageMaker.pdf
dynamic_feat (optional)—An array of arrays of floating-point values or integers that represents the vector of custom feature time series (dynamic features). If you set this field, all records must have the same number of inner arrays (the same number of feature time series). In addition, each inner array must have the same length as the associated target value. Missing values are not supported in the features. For example, if target time series represents the demand of different products, an associated dynamic_feat might be a boolean time-series which indicates whether a promotion was applied (1) to the particular product or not (0):
not mentioned type of possible values ??
can integer represent categorical? or dynamic_feat may have only numerical values?
Think of cat as static features of that time series. For example, in the Electricity dataset example, we have several time series, one for each "house". In a time series for a specific house, we could have some categories like the zipcode, or house size in sq. ft. Those are static features, because they don't change over time for that specific time series.
On the other hand, dynamic_feat change over time. Let's say we want to have the average temperature associated with each datapoint. The avg_temps would be inside dynamic_feat. That's why the length of target and the length of each inner array in dynamic_feat must be equal.
cat must be integer, sequentially encoded.
dynamic_feat must be numerically encoded.
