1-d infill error estimate

[achase90]

I think there might be a bug in the kriging or MSE estimate if the problem is compressed to 2-D (ie, one input and one output). If you run the x*sin(x) test function with only a few points (2-4) and plot the error estimate, it doesn't look correct.

Is the smoothness factor "p" getting optimized on as well as the theta values? If not, that might cause this issue. I believe this bug also exists in the Matlab code, so that won't be a great resource to compare against.

[capaulson]

I'll take a look. 1-d arrays weren't really tested during development. I need to add some checks to make sure that the array is at least 2-d. The p factor is optimised in this code. Could you post a quick script to recreate the issue you're seeing with x*sin(x)?

[hchasestevens]

I've noticed something similar in working with 1-d arrays - please see the code below and resultant output.

import pyKriging
from pyKriging.krige import kriging

MAXVAL = 32
CONVERSION = 0.9687491103216275

X = np.array([[0.01], [1], [2], [8], [16], [32]]) / MAXVAL
y = np.array([
    4.97292791842937,   
    4.84888042822178,   
    4.843428519136497,  
    4.885980222410153,  
    4.952942447020513,  
    5.06074067099633,   
]) * CONVERSION

k = kriging(X, y)
k.train(optimizer='pso')
k.snapshot()

from matplotlib import pyplot as pl
pl.scatter(X * MAXVAL, y, c='black', s=20)
xs = [x / 1000. * MAXVAL for x in xrange(1000)]
k_predictions = [k.predict([x / MAXVAL]) for x in xs]
pl.plot(xs, k_predictions,)
errors = [k.predict_var([x / MAXVAL]) for x in xs]
pl.fill(
    xs + xs[::-1], 
    [pred + err for pred, err in zip(k_predictions, errors)] +
    [pred - err for pred, err in zip(k_predictions, errors)][::-1],
    alpha=.25,
    fc='black'
)
pl.show()

This results in:

Any assistance would be greatly appreciated.

[capaulson]

Hi, thanks for the example. I'm looking into this now. It seems to be an error in how the predictions are generated. More to come soon.

Best, Chris

[achase90]

Did you ever figure out what was going on with this?

[hchasestevens]

Hi @capaulson , just wondering if you ever determined the cause of the issue here. Thanks!