有没有办法在 Numpy 中实现更新的伪随机数生成器

Question

I would like to try different pseudo-random number generators (PRNG) in Numpy. It is mostly an academic curiosity to see how different algorithms compare. Does Numpy has a straightforward way of pointing to a different PRNG, or would I need to alter the source code?

Answer 1

Changing the random number generation is not a provided method / thing you can do with python, you can change the seed for the generation however, but this will do little for the actual generation of random numbers, Python uses the Mersenne Twister algortihm to generate its random numbers, if you want to change the algorithm you will need to change the source code for it. I don't know where it is, but I did do some research into this stuff a while back. Heres a wiki link for all the random numnber generation algorithms, I dont recommend coding any of them as they are ment to be complex, hinting towards the idea of randomness. link:

https://en.wikipedia.org/wiki/List_of_random_number_generators

In addition, python does offer some variety in the forms of generation, however the random method is pretty stedfast for whatever you may want to use it for, if you want a faster method you can go with a TinyMT but you will again need to code it by hand. Below is a link to all of the random methods, including the seed:

https://docs.python.org/3/library/random.html

Hope this helped, and I would advise against trying to change the source code for random number generation, as randomness only really becomes a factor in security and otherwise a basic system like MT will do fine.