scivislab/bdi-ed

Branch Decomposition-Independent Edit Distances

bdi-ed's Issues

Surprising Output on Interval-Based Merge Tree

Hello,
This code seems really neat, but I either stumbled across a bug or I'm not understanding how to use it properly.

I setup a very simple experiment for two merge trees on the interval, one which is exactly the same but a leaf node was moved up by 1. Everything else is exactly the same, including the topology of the trees. I would expect the distance to be proportional to 1, but I'm seeing only a distance of 0. Picture of the two trees is shown below:

Here is the code I ran to create this example. Am I making a mistake on how I ran this? I would have expected the distance to be greater than 0.
Thank you for your help!
Sincerely,
Chris Tralie

    import numpy as np
    import matplotlib.pyplot as plt
    from baseMetrics import *
    from mergeTreeEdit_branch import *
    
    def plot_tree(idx, nodeScalars, children, x=0, rg=2):
        y = nodeScalars[idx]
        plt.scatter(x, y)
        plt.text(x+0.03, y-0.01, "{}".format(idx))
        N = len(children[idx])
        for i, c in enumerate(children[idx]):
            x2 = x-rg/2+i*rg/(N-1)
            xs = [x, x2]
            ys = [y, nodeScalars[c]]
            plt.plot(xs, ys, 'k')
            plot_tree(c, nodeScalars, children, x2, rg/2)
    
    nodeScalars1 = [0, 5, 3, 4, 2]
    children1 = [[], [0, 3], [], [2, 4], []]
    rootID1 = 1
    
    nodeScalars2 = [0, 5, 2, 4, 1]
    children2 = [[], [0, 3], [], [2, 4], []]
    rootID2 = 1
    
    plt.figure(figsize=(8, 3))
    plt.subplot(121)
    plot_tree(rootID1, nodeScalars1, children1)
    plt.title("Tree 1")
    plt.subplot(122)
    plot_tree(rootID2, nodeScalars2, children2)
    plt.title("Tree 2")
    plt.show()
    
    
    dist = branchMappingDistance(nodeScalars1,children1,rootID1,nodeScalars2,children2,rootID2,cost_wasserstein_branch)
    print(dist)