JointRecurrencePlot return zero

[Rariwa]

Hi, I have 5 features and 100 timesteps. JoinRecurrentPlot return all values equal to zeroes. is there anything wrong in my data?

[Johann Faouzi]

Hi,

A joint recurrence plot is the Hadamard (element-wise) product of all the individual recurrence plots (one for each feature). Thus, for each "pixel", a zero for at least one feature will result in a zero in the joint recurrence plot. To tackle this, you can decrease the number of zeros in any recurrence plot by increasing the threshold used to binarize the distance between the trajectories.

Here is an example on a toy dataset:

import matplotlib.pyplot as plt
from pyts.datasets import load_basic_motions
from pyts.multivariate.image import JointRecurrencePlot

X, _, _, _ = load_basic_motions(return_X_y=True)

With a small number of ones in each recurrence plot, there are almost only zeros in the joint recurrence plot:

jrp_1 = JointRecurrencePlot(threshold='point', percentage=5)
X_jrp_1 = jrp_1.transform(X)
plt.imshow(X_jrp_1[0], cmap='binary', origin='lower')

You can increase the number of ones in each recurrence plot, which results in many more ones in the joint recurrence plot:

jrp_2 = JointRecurrencePlot(threshold='point', percentage=50)
X_jrp_2 = jrp_2.transform(X)
plt.imshow(X_jrp_2[0], cmap='binary', origin='lower')

You can also specify different thresholds for different features:

jrp_3 = JointRecurrencePlot(threshold='point', percentage=[50, 10, 70, 60, 50, 40])
X_jrp_3 = jrp_3.transform(X)
plt.imshow(X_jrp_3[0], cmap='binary', origin='lower')

Hope this helps you a bit.