When r=d=500, it does not work

[XingLiu93]

Hi, I met a problem when I use the program. When I test with 500 elements on r and d, it always has error:
ValueError: setting an array element with a sequence.
But with the same data, when len(r)=len(d)=20, it works. Do you know why?
This is my questions. https://stackoverflow.com/questions/46270121/python-program-error-valueerror-setting-an-array-element-with-a-sequence
Thanks

[Wolfi-101]

For me, the "setting an array with a sequence" error came up if the minimum of the cost function occurred more than once, i.e. the line np.where(cost[t,:] == min(cost[t,:])) returned an array with more than one entry. I know of two situations where this can happen:

1. The probability for both states is the same. This happens if s=1. At the first timestep, the fit function will return the same value for both states. To avoid this, assert that s>1.
2. The values for r and d are too high. In the fit function, the value for the binomial coefficient and the probabilities p^r and (1-p)^(d-r) are computed separately. If d is very large, for example d>100000, the binomial coefficient will be np.inf or the whole thing will evaluate to np.nan. I chose to use scipy.stats.binom.pmf(k=r, n=d, p=p) which avoids these separate calculations and returns the value of the probability mass function straight away.

Hope this helps

[Hsgngr]

I have the same issue on my Windows while my colleague didn't get any error in his mac.