Jupyter

To get these examples to work in Jupyter, you will need to install the following.

• ipywidgets

• ffmpeg

Widgets

Let’s see how we can use zava to work with ipywidgets. First, we got to get some data.

import numpy as np
import pandas as pd

M = np.array([
    [1, 1, 1, 1],
    [2, 2, 2, 1],
    [3, 3, 3, 3],
    [1, 2, 3, 4],
    [2, 2, 1, 1],
    [1, 1, 3, 3]
])

Now we create an instance of GrandTour with the data and also specifying the minimum c and maximum d values for scaling.

from zava.core import GrandTour

c = 0
d = 1
grand_tour = GrandTour(M, c, d)

Finally, we use a function f annotated with @interact to create an interactive visualization with parallel coordinates and Grand Tour.

import matplotlib.pyplot as plt
from ipywidgets import interact

@interact(degree=(0, 360 * 4, 0.5))
def f(degree=0):
    S = grand_tour.rotate(degree)

    fig, ax = plt.subplots(figsize=(15, 3))

    params = {
        'kind': 'line',
        'ax': ax,
        'color': 'r',
        'marker': 'h',
        'markeredgewidth': 1,
        'markersize': 5,
        'linewidth': 0.8
    }

    _ = S.plot(**params)
    _ = ax.get_legend().remove()
    _ = ax.set_xticks(np.arange(len(S.index)))
    _ = ax.set_xticklabels(S.index)
    _ = ax.get_yaxis().set_ticks([])
    _ = ax.set_title('Grand Tour')

Animations

Now let’s see how we can create HTML5 animations in a notebook using matplotlib.animation. Again, start with some data.

import numpy as np
import pandas as pd

M = np.array([
    [1, 1, 1, 1],
    [2, 2, 2, 1],
    [3, 3, 3, 3],
    [1, 2, 3, 4],
    [2, 2, 1, 1],
    [1, 1, 3, 3]
])

Create a GrandTour instance with the data.

from zava.core import GrandTour

c = 0
d = 1
grand_tour = GrandTour(M, c, d)

We have to wrap the GrandTour instance with a SinglePlotter. The SinglePlotter plots only a single set of data with an axis and does not concern itself with the greater plot (e.g. the title). The params argument is a dictionary that you can override to change the line drawings.

from zava.plot import SinglePlotter

single_plotter = SinglePlotter(grand_tour, params={'color': 'r'})

The MultiPlotter controls all the plots and takes in a list of SinglePlotters as well as an axis. You can then use an instance of this object with animation.FuncAnimation() as usual to produce an animation.

from zava.plot import MultiPlotter
from matplotlib import animation

fig, ax = plt.subplots(figsize=(5, 3))

multi_plotter = MultiPlotter([single_plotter], ax=ax)

params = {
    'fig': fig,
    'func': multi_plotter,
    'frames': np.linspace(0, 360, 360),
    'interval': 20,
    'init_func': multi_plotter.init
}
anim = animation.FuncAnimation(**params)

plt.close(fig)

Finally, render the video.

%%time

from IPython.display import HTML

HTML(anim.to_html5_video())

CPU times: user 21.4 s, sys: 574 ms, total: 22 s
Wall time: 22.1 s

Animation, colors

You might find yourself doing cluster analysis of high-dimensional data. If you recover some clusters, you can break the data apart according to the clusters and visualize them with different colors. Here’s a full working example (without the clustering) of how to visualize two datasets.

%%time

# 1. here are your two datasets, M1 and M2

columns = ['v0', 'v1', 'v2', 'v3']

M1 = np.array([
    [1, 1, 1, 1],
    [2, 2, 2, 1],
    [3, 3, 3, 3]
])
M2 = np.array([
    [1, 2, 3, 4],
    [2, 2, 1, 1],
    [1, 1, 3, 3]
])

M1 = pd.DataFrame(M1, columns=columns)
M2 = pd.DataFrame(M2, columns=columns)

# 2. create your GrandTour instances

c = 0.0
d = 100.0

gt1 = GrandTour(M1, c, d)
gt2 = GrandTour(M2, c, d)

# 3. create corresponding SinglePlotters

sp1 = SinglePlotter(gt1, params={'color': 'r'})
sp2 = SinglePlotter(gt2, params={'color': 'g'})

# 4. create a MultiPlotter from the SinglePlotters
fig, ax = plt.subplots(figsize=(5, 3))
mp = MultiPlotter([sp1, sp2], ax=ax)

params = {
    'fig': fig,
    'func': mp,
    'frames': np.linspace(0, 360, 360),
    'interval': 20,
    'init_func': mp.init
}
anim = animation.FuncAnimation(**params)

plt.close(fig)

# 5. display the animation
HTML(anim.to_html5_video())

CPU times: user 25 s, sys: 608 ms, total: 25.6 s
Wall time: 25.7 s