Utilities for randomly sampling from statistical distributions

Simple Random Distribution Sampling

SRDS is mainly a wrapper around scipy's statistical functions (scipy.stats). It makes it easier to sample from parameterized distributions and provides tools that accelerate random sampling.

Examples

Truncation or rejection sampling

srds adds several classes that make it easier to utilize scipy statistical distributions. To sample from a log-normal distribution with `σ = 0.25`, but truncate at `0.5` and `2`, the `BoundRejectionSampler` helps:

```from srds import ParameterizedDistribution as PDist, BoundRejectionSampler

dist = PDist.lognorm(0.25)
sampler = BoundRejectionSampler(dist, 0.5, 2)

x = sampler.sample(10)
```

Fast sampling of single values

calling `dist.rvs` on a scipy statistical distribution is computationally expensive. This is problematic for code that does something like:

```# will be slow (calls dist.rvs 10000 times)
for i in range(10000):
x = dist.sample()
# ...
```

srds provides a `BufferedSampler` that draws a larger sample from a distribution, and subsequently returns from that sample.

```from srds import BufferedSampler

dist = BufferedSampler(dist)

# will be much faster! (calls dist.rvs only 10 times with a sample size of 1k)
for i in range(10000):
x = dist.sample()
# ...
```

Sampling from populations

You can use the `PopulationSampler` to draw from a discrete set, and also bias the sampling with weights.

```from srds import PopulationSampler

sampler = PopulationSampler(['a', 'b', 'c'], [8, 1, 1])
sampler.sample() # will return 'a' 8 out of 10 times on average
sampler.sample(10) # returns a list containing items from ['a', 'b', 'c'] in random order
```

Project details

Source Distribution

srds-0.0.1.dev1.tar.gz (6.1 kB view hashes)

Uploaded `source`

Built Distributions

srds-0.0.1.dev1-py3.7.egg (13.3 kB view hashes)

Uploaded `3` `7`

srds-0.0.1.dev1-py3-none-any.whl (7.6 kB view hashes)

Uploaded `py3`