Is it not possible to apply numpy in a cell?

[bakerwy]

Hi,

I am very intrigued by vectorized approach.
As you know, actuarial models use uncountable 'if .. elif.. else' sentences and it is little bit tricky for vectorized model.

I found numpy has useful function for that and it looks helpful for vectorized modeling.

For example, I can use the code below which is easy to understand

def NO_SURRS(t):
conditions = [t > POL_TERM_M(),
t <= DUR_M(),
t <= PREM_PAYBL_M()]
result = [np.zeros_like(t),
NO_POLS_IFSM() * LAPSE_RATE_DEPNDT(),
NO_PAY_IFSM() * LAPSE_RATE_DEPNDT()]

return np.select(conditions, result, default = NO_POLS_IFSM() * LAPSE_RATE_DEPNDT())

but when I tried this function, the error message pops up and it said np is not defined even if I wrote Import Numpy as np at the top of the code.

Instead of previous code, the code below works well.

def NO_SURRS(t):
import numpy as np
conditions = [t > POL_TERM_M(),
t <= DUR_M(),
t <= PREM_PAYBL_M()]
result = [np.zeros_like(t),
NO_POLS_IFSM() * LAPSE_RATE_DEPNDT(),
NO_PAY_IFSM() * LAPSE_RATE_DEPNDT()]

return np.select(conditions, result, default = NO_POLS_IFSM() * LAPSE_RATE_DEPNDT())

The problem is that the models will have many cells use that method and each of cells need to call import numpy as np.
Also, in my knowledge, Modelx uses numpy.

Is there any way to call numpy once?

[fumitoh]

Yes. If NO_SURRS is defined in the Projection space, then you can assign the numpy module to the np attribute of Projection like this:

import numpy as np
Projection.np = np

After this, np becomes referable in any formulas in Projection, including NO_SURRS.

Insted, if you assign the numpy module to the np attribute of the model instead, then np becomes referable from any formulas in all spaces in the model.