Basic models

You can use analytic probability distributions and histograms as basic models, described below.

Distributions

Hypney supports the univariate probability distributions from scipy.stats (https://docs.scipy.org/doc/scipy/reference/stats.html, accessed by the same name:

import hypney.all as hp

hp.norm().plot_pdf()
hp.cosine().plot_pdf()
hp.cauchy().plot_pdf(x=np.linspace(-5, 5, 500))

(Source code, png, hires.png, pdf)

Discrete distributions are supported too:

hp.poisson(mu=3).plot_pdf()
hp.binom(n=8, p=0.2).plot_pdf()

(Source code, png, hires.png, pdf)

Hypney uses scipy.stats for computations on numpy data. For tensorflow data, it will use corresponding distributions from tensorflow-probability (https://www.tensorflow.org/probability/api_docs/python/tfp/distributions) (if available), and similarly torch.distributions (https://pytorch.org/docs/stable/distributions.html) for PyTorch data and jax.scipy.stats(https://jax.readthedocs.io/en/latest/jax.scipy.html#jax-scipy-stats) for JAX data.

Hypney assumes all shape parameters have a default of 0, and that the distributions are implemented consistently between scipy, tensorflow-probability, torch, and JAX. We should probably test that explicitly soon…

Histograms

You can also build a model from a histogram:

` hp.from_histogram... `