SALib/SALib

Sensitivity Analysis Library in Python. Contains Sobol, Morris, FAST, and other methods.

What it solves

SALib helps you understand how the inputs of a system model affect its outputs. When building complex models, it is often unclear which input variables are the most influential or how they interact with each other. This library provides the tools to quantify these relationships, helping you identify which factors drive your model's results.

How it works

SALib provides a collection of established sensitivity analysis methods. You define your model's parameters (names and bounds), and the library generates samples for those parameters. You run your model using these samples, and then pass the results back to SALib to perform the analysis. It supports both a procedural approach for step-by-step control and a chained approach for a more streamlined workflow.

Who it’s for

It is designed for researchers, data scientists, and engineers who build systems models and need to perform global sensitivity analysis to validate or interpret their work.

Highlights

  • Includes a wide range of sensitivity analysis methods such as Sobol, Morris, eFAST, and Shapley Effects.
  • Supports both command-line and Python-based workflows.
  • Offers a flexible 'ProblemSpec' interface for chaining sampling, evaluation, and analysis steps.
  • Provides built-in plotting functionality to visualize analysis results.
  • Compatible with standard Python data science tools like NumPy, SciPy, and pandas.

Related

  • Project
    yueliu1999/Awesome-Deep-Graph-Clustering

    ADGC is a curated list of papers, code links, and datasets for deep graph clustering, organized by sub‑topics such as LLM‑based methods, temporal models, and approaches that don’t need a predefined number of clusters.

  • Project
    probml/dynamax

    A JAX-based library for probabilistic state space models that provides tools for inference and learning across various model types like HMMs and Gaussian SSMs.

  • Project
    YingfanWang/PaCMAP

    PaCMAP is a dimensionality reduction tool that creates high-quality data visualizations by simultaneously preserving both local and global structures. It offers a scikit-learn compatible interface and supports large-scale datasets through various optimized neighbor-search backends.

  • Project
    alex-petrenko/sample-factory

    A high-throughput reinforcement learning library focused on efficient synchronous and asynchronous implementations of PPO to minimize training time andes hardware requirements.