ximinf

Simulation-Based Inference of Cosmological Parameters in JAX using Type Ia Supernovae.

ximinf is a Python package for performing simulation-based inference (SBI) on cosmological parameters using type Ia supernova (SN Ia) data, built on top of JAX for fast, differentiable, and GPU/TPU-accelerated computation.

Installation

pip install ximinf

Requires Python >= 3.10.

Quickstart

import ximinf

See the documentation for full usage examples covering simulation, training, and inference.

Features

ximinf is organized into five core modules:

ximinf.generate_sim — Generates simulated SN Ia datasets. Includes Latin Hypercube Sampling (LHS) of cosmological/nuisance parameters under configurable priors (uniform, gaussian, half-gaussian, log-uniform, exponential, etc.), and per-dataset simulation via skysurvey.
ximinf.selection_effects — Tools for injecting realistic observational selection effects (e.g. Malmquist bias) into simulated supernova samples via stochastic magnitude-limited detection.
ximinf.nn_train — Training utilities for the neural network classifiers used in the inference step, including loss/accuracy functions, training/validation loops with early stopping, and JAX device setup helpers.
ximinf.nn_test — Diagnostics for validating inference quality, notably TARP (Tests of Accuracy with Random Points) coverage statistics computed across parameter groups.
ximinf.nn_inference — Neural posterior/likelihood-ratio inference. Builds and runs BlackJAX NUTS samplers over grouped parameter blocks, combining trained neural network outputs with analytic log-priors.

Typical workflow

Simulate — Draw parameter sets with generate_sim.scan_params and generate mock SN Ia samples with generate_sim.simulate_one.
(Optional) Apply selection effects — Use selection_effects.apply_malmquist_bias to emulate realistic detection thresholds.
Train — Train per-group neural classifiers on simulated data with nn_train.train_loop.
Infer — Sample the posterior over cosmological parameters with nn_inference.sample_posterior / nn_inference.inference_loop, using a NUTS kernel built on the trained networks.
Validate — Check calibration of the resulting posteriors with nn_test.compute_ecp_tarp_groups.

License

This project is licensed under the GNU General Public License v3.0 (GPLv3). See LICENSE for details.

Author

Adam Trigui — a.trigui@ip2i.in2p3.fr IP2I Lyon

Release files for ximinf 0.0.157

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ximinf 0.0.157
File	Size	Uploaded
ximinf-0.0.157.tar.gz	41.1 kB	Sep 22, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for ximinf 0.0.157
File	Interpreter	ABI	Platform	Reset
ximinf-0.0.157-py3-none-any.whl 43.5 kB Sep 22, 2026	Python 3	none	any	Details

Total release size: 84.6 kB

Release files / ximinf-0.0.157.tar.gz

Download URL	ximinf-0.0.157.tar.gz
Size	41.1 kB
Tags	Source
SHA-256 checksum How to use checksums	`8d27f47d940c1d8d60b87c2626effd9d6cdef4840d52b87210469ad2d4fbe123`
BLAKE2b-256 checksum How to use checksums	`79fe0ab8b30dd07de2098b397e28495101847459dc356d1393de44faa350b44d`
Upload date	Sep 22, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.11.16`

Release files / ximinf-0.0.157-py3-none-any.whl

Download URL	ximinf-0.0.157-py3-none-any.whl
Size	43.5 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`75fba03f94f7c6843fe514203cf1f83bea738060f6943ae4e5d38488166143df`
BLAKE2b-256 checksum How to use checksums	`c468abc0ad96cdb181d760faf9f27ac0313caf9833b809dcd177a310610d0db3`
Upload date	Sep 22, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.11.16`