Modularised plotting about agentmet4fof HOT 9 CLOSED

Maybe mpld3 can help?

from agentmet4fof.

For the matplotlib:

• branch out the options to choose 1. or 2.
• commit the progress with mpld3 and laeve it as draft, so that it will be tracked

from agentmet4fof.

For the scenarios of custom plots:

• Document the options of 1. sending plot figures, or 2. plot within the monitor agents' memory
• maintain these two options
• write tutorials on these

from agentmet4fof.

Thanks for improving the clarity!

from agentmet4fof.

There are 3 ways to include matplotlib figures in the web dashboard powered by plotly:

1. convert to static image (existing method: easiest and exact but no interactivity)
2. convert to plotly figure via tls.mpl_to_plotly (fully interactive, but not all matplotlib figures are compatible such as boxplots, piecharts,..)
3. mpld3. This requires some hack to work with current dashboard, but making good progress.. (sufficient interactivity, wide compatibility with most matplotlib figures)

perhaps then, there should be an option for users to switch from 1 method to the other.

from agentmet4fof.

In agentMET4FOF context, there are probably 3 scenarios which custom plots could take place:

1. Send plot figures from data processing/streaming agents to Monitor agent
2. Plotting function supplied with Monitor agent's init_parameters, and actual plotting is based on Monitor agent's memory data
3. Offline - ML experiments for comparing results of ML pipelines performances/outputs

from agentmet4fof.

For the matplotlib:

* branch out the options to choose 1. or 2.

* commit the progress with mpld3 and laeve it as draft, so that it will be tracked

I would describe my suggestion with the words:

• push the according branch (dealing with all three options, as far as I understood)
• create a Pull Request for this branch
• either revert the commits introducing mpld3 if they were confined to the third option or just remove the according snippets from the code in one commit with a self-explaining message telling clearly when we dropped the changes related to its use
• merge the Pull Request after properly reviewing the left-overs and desired changes

This way we can properly proceed with this issue, resolve it and we do not loose the first try of introducing mpld3, which was promising.

from agentmet4fof.

An example of how to provide custom plot of plotly is provided here:

https://github.com/bangxiangyong/agentMET4FOF/blob/develop/examples/custom_dashboard.py

from agentmet4fof.

addressed here, within the send_plot function of base agent class:

https://github.com/bangxiangyong/agentMET4FOF/blob/custom_plots/agentMET4FOF/agents.py

from agentmet4fof.