Open andrewheusser opened 7 years ago
Ideally data streams would be just another type of supported data types, in addition to pandas dataframes and numpy arrays.
So we'd use some initial samples to compute the dimensionality reduction transform, and then all future samples would be added to the plot dynamically using that transform.
(so I'd vote for no streaming flag)
Potentially useful library (thanks for the twitter tip, @andrewheusser!): http://streamz.readthedocs.io/en/latest/
A tutorial-style description may be found here: https://matthewrocklin.com/blog//work/2017/10/16/streaming-dataframes-1
This might be useful for adding support for streaming data: https://ipywidgets.readthedocs.io/en/latest/
e.g.
hyp.plot(streaming_data, streaming=True)
or could be inferred from the structure of
streaming_data