numpyro

numpyro copied to clipboard

Customize MCMC algorithm

4

Hi all, thanks for your powerful library. Sorry for my beginner question. Now I have implemented a Metropolis-Hasting algorithm following this [link](https://num.pyro.ai/en/stable/mcmc.html#numpyro.infer.mcmc.MCMCKernel), I need to customize the score calculation. My...

DQSSSSS

I would like to do batch transformations using `transforms.LowerCholeskyAffine`. So far I can do a "batched" transform as long as the `loc` and `scale_tril` are the same however if I...

chrisfinlay

GMM notebook example: MCMC/NUTS simulation is not reproducible

2

``` source: https://num.pyro.ai/en/stable/tutorials/gmm.html#MCMC numpyro.__version__: 0.12.1 jax.__version__: 0.4.13 ``` -- When running the collapsed NUTS to explore the full posterior, the obtained results did not match the presented ones. ```python from...

inarighas

Hi all, I'm trying to initialize my MCMC with one or more specific values for each chain. As numpyro.infer.util.init_to_value() doesn’t work for multiple chains and init_params argument takes only unconstrained...

asigot

init_strategy does not work with CircularReparam

1

Hi! I am having an issue with initialising a model with the init_by_median() strategy for my MCMC NUTS model. I am drawing the location parameter of a VonMises distribution from...

19MM97

Similar to `render_model` but render to a LaTex formula. Example: ```py def model(data): m = pyro.sample("m", dist.Normal(0, 1)) sd = pyro.sample("sd", dist.LogNormal(m, 1)) with pyro.plate("N", len(data)): pyro.sample("obs", dist.Normal(m, sd), obs=data)...

ordabayevy

Different supports in component distributions for mixture models

1

Fixes #1780.

Qazalbash

Add HSGP contribution module

6

Closes https://github.com/pyro-ppl/numpyro/issues/1788 **Motivation:** As described in the issue > I have been using the Hilber Space approximation for Gaussian Processes a lot. The main ingredients are the spectral densities of...

juanitorduz

This PR adds the Wishart distribution for covariance matrices. It should probably be rebased and reviewed after #1778 is merged.

tillahoffmann

HSGP utility functions in the `contrib` module?

2

I have been using the Hilbert Space approximation for Gaussian Processes a lot. The main ingredients are the spectral densities of the kernels and the eigenvalues and eigenvectors of the...

juanitorduz