mergeron.core.pseudorandom_numbers¶
Functions for generating synthetic data under specified distributions.
Uses multiple CPUs when available, with PCG64DXSM as the PRNG. [1]
References
Module Contents¶
-
mergeron.core.pseudorandom_numbers.prng(_s=
None, /)[source]¶ Return a psure-random number generator.
-
mergeron.core.pseudorandom_numbers.seed_sequencer(_len=
3, /, *, generated_entropy=None)[source]¶ Return specified number of SeedSequences, for generating random variates.
Initializes a specified number of SeedSequences based on a set of 10 generated “seeds” in a hard-coded list. If the required number of random variates is larger than 10, the user must first generate a sufficient number of seeds to draw upon for initializing SeedSequences. The generated entropy can be reused in subsequent calls to this function.
- Parameters:¶
- Returns:¶
A list of numpy SeedSequence objects, which can be used to seed prng() or to spawn seed sequences that can be used as seeds to generate non-overlapping streams in parallel. [2]
- Raises:¶
ValueError – When, \(\_sseq\_list\_len > max(10, len(generated\_entropy))\).
- Return type:¶
tuple[numpy.random.SeedSequence, Ellipsis]
References
- class mergeron.core.pseudorandom_numbers.MultithreadedRNG[source]¶
Fill given array on demand with pseudo-random numbers as specified.
Random number generation is multithreaded, using twice the number of threads as available CPU cores by default. If a seed sequence is provided, it is used in a thread-safe way to generate repeatable i.i.d. draws. All arguments are validated before commencing multithreaded random number generation.
- values : mergeron.ArrayDouble[source]¶
Output array to which generated data are over-written
Array-length defines the number of i.i.d. (vector) draws.
- dist_type : Literal['Beta', 'Dirichlet', 'Gaussian', 'Normal', 'Random', 'Uniform'][source]¶
Distribution for the generated random numbers.
Default is “Uniform”.
- dist_parms : mergeron.ArrayFloat[source]¶
Parameters, if any, for tailoring random number generation