Investigate possible lossy aggregations

[DylanVanAssche]

To save bandwidth, we could have an tss:SampledSnippet which provides a resampling based on a certain strategy.

[DylanVanAssche]

From @pietercolpaert : we need features in the snippets to create aggregated versions on the fly.

[pietercolpaert]

Might be interesting to investigate https://github.com/DynamicsAndNeuralSystems/catch22

[DylanVanAssche]

Now that we are switched to JSON as format to store the data points, we can easily re-use existing tooling like Catch22:

https://time-series-features.gitbook.io/catch22/language-specific-docs/python#getting-started-basic-usage

It requires an array of values, which we can easily extract with JSONPath now:

$.[*].value

[DylanVanAssche]

Note that catch22 only deals with:

single univariate time series as input

due to processing overhead of variating time series and multiple ones.

[pietercolpaert]

Note that catch22 only deals with single univariate time series as input

Yes, I think this aligns very well with what we want to achieve: generic features of this time series snippet