Numpy returns an overflow error when handling large numbers

Asked Oct 16 '21 at 10:00

Active Oct 16 '21 at 10:00

Viewed 28 times

Let me explain.

So I'm running a simulation with 2^15 points in MATLAB. Part of my initial condition is to use the

Sech().^2

Function. This works perfectly fine in MATLAB and I get an array of very useable data.

In python on the other hand, when I go

1/( np.cosh(x)**2 )

I get a runtime/ overflow error that ruins my day.

Why can MATLAB handle the code but numpy can't?

RuntimeWarning: overflow encountered in cosh r = 1/(np.cosh(x)**2)

The code works flawlessly when translated in MATLAB and produces exactly what I want.

asked Oct 16 '21 at 10:00

Dimitri_896

Because numpy used C-ish types that can overflow. Make sure you use the right type for your array. – Klaus D. Oct 16 '21 at 10:05
What is the correct type of array here? I'm a bit confused – Dimitri_896 Oct 16 '21 at 10:10
repeat of https://stackoverflow.com/questions/69590932/runtimewarning-overflow-encountered-in-cosh-python-what-does-this-mean. @Klaus numpy is **warning** that the `cosh` returns `inf` – hpaulj Oct 16 '21 at 10:10
It's a repeat yes but I'm curious to know what the correct way to do this is. It works in MATLAB – Dimitri_896 Oct 16 '21 at 10:14
overflow warnings can be suppressed. MATLAB has the `sech` which goes to 0, while in numpy you have to use `cosh` as an intermediate which goes to `inf` – hpaulj Oct 16 '21 at 10:25
Do you know how? – Dimitri_896 Oct 16 '21 at 10:27
I added a warning suppress example to my previous answer. – hpaulj Oct 16 '21 at 16:11

0 Answers0