[ENH] 2d Array Plotting

[ryanhammonds]

This updates plot_time_series to accept 2d arrays, in addition to a list of 1d arrays. Before this would not work:

import numpy as np
from neurodsp.plts import plot_time_series

n_sigs = 10
fs = 1000
n_seconds = 10

sigs = np.repeat(np.arange(n_sigs), int(n_seconds * fs))
sigs = sigs.reshape((n_sigs, int(n_seconds * fs)))

times = np.arange(0, len(sigs[0])/fs, 1/fs)

plot_time_series(times, sigs)

and it would require:

plot_time_series(times, list(sigs))

[TomDonoghue]

Yeh, this is a totally sensible update. I've made some review comments for condensing etc. Also: maybe extend the test to test the case of having a 2d array input? Also, there are some 'secondary' functions which use plot_time_series (such as plot_bursts and plot_instantaneous_measure - can you have a quick check that this update doesn't induce anything weird with their outputs?

Here is a slightly different example in which this update does help to plot things cleanly:

from neurodsp.sim import sim_oscillation
from neurodsp.plts import plot_time_series
from neurodsp.utils import create_times

n_seconds, fs = 2, 1000

powers = [0, 0.25, 0.5, 0.75, 1]

times = create_times(n_seconds, fs)
sigs = np.array([sim_oscillation(n_seconds, fs, 10, variance=power) for power in powers])

plot_time_series(times, sigs)

Output:

This is a slightly different from the notion of plotting something like 'multiple channels':

from neurodsp.sim import sim_oscillation
from neurodsp.plts import plot_time_series
from neurodsp.utils import create_times

n_seconds, fs = 2, 1000
powers = [0, 0.25, 0.5, 0.75, 1]

times = create_times(n_seconds, fs)
sigs = np.array([sim_oscillation(n_seconds, fs, 10, variance=power) + 2*ind for ind, power in enumerate(powers)])

plot_time_series(times, sigs)

output:

Note, for example, that the y-axis labels and ticks don't really make sense in this case.

The point being: I think we should merge this update in for the current function, but this also perhaps highlights that we should add a different / new function for doing 'multi-channel'? This feels like it could be useful, though perhaps somewhat out of scope?

[ryanhammonds]

Thanks for the review! I went a little wild with trying to avoid to use repeat. I've incorporated your feedback.

This plays nicely with plot_instantaneous_measure. And doesn't change anything with plot_bursts - it only took a 1d array and that hasn't changed. A list or 2d array will break plot_bursts. I could change that if you thinks it's useful. I'll take some playing around with since vstacking sig and bursts doesn't work.

I think a multi-channel plotting function would be useful and wouldn't be too much work to write in but might overlap with MNE. I took a stab at this in ndspflow with plotly - it ended up not working super well for large arrays there since there was too much interactivity and something deep in plotly's codebase/dependencies caused slow loading and poor performance for what I was trying to do.

[TomDonoghue]

If this doesn't change anything with instaneous, and bursts, I think we're good. I don't think it's an issue that bursts doesn't accept 2d - it's not even clear what it should do in that case, and it's documented to only take in a 1d array.

With that, the code & update in this PR all looks good to me, you can merge it in @ryanhammonds!

In terms of the multi-channel plotting (as a potential separate thing, not for this PR) - I agree, it could be useful, and try it out if you want - but it's not a strong priority necessarily.