pymc-devs/pymc
Pymc
PyMC is a Bayesian probabilistic programming framework used for building probabilistic models and performing Bayesian inference. It provides a probabilistic graphical model library for specifying random variables, priors, and likelihood functions, supported by an MCMC sampling engine and variational inference tools to estimate posterior distributions.
The framework features a GPU-accelerated inference backend that compiles models into machine code to increase execution speed. It utilizes a backend-agnostic tensor execution model and just-in-time graph compilation to optimize the computation of log-probabilities and gradients.
The project covers a wide range of statistical modeling capabilities, including Gaussian processes, survival analysis, causal inference, and time series forecasting. It supports the construction of generalized linear models, mixture models, and the integration of ordinary differential equations within probabilistic workflows.
The system includes tools for model convergence diagnosis and posterior distribution analysis to evaluate inference quality and model fit.
Features
- Markov Chain Monte Carlo Sampling - Implements Markov Chain Monte Carlo algorithms to sample from complex posterior distributions.
- Probabilistic Modeling - Provides a comprehensive framework for constructing probabilistic models using random variables to compute joint log-probabilities and gradients.
- Bayesian Statistical Modeling - Provides a comprehensive framework for defining probabilistic models with priors and likelihoods to perform Bayesian inference.
- Probabilistic Generative Sampling - Generates random samples from joint probability distributions using various computational backends.
- Likelihood Models - Constructs likelihood models that represent observed data for use in Bayesian inference.
- Hardware Acceleration - Compiles models into machine code to run graph operations and inference on GPUs.
- Parameter Estimation Methods - Provides statistical methods for estimating unknown model parameters from observed data.
- Posterior Inference - Implements posterior inference methods to update prior distributions with observed data for uncertainty quantification.
- Variational Inference Implementations - Implements variational inference techniques to approximate complex posterior distributions via optimization.
- Probabilistic Graphical Models - Provides a probabilistic graphical model library for specifying random variables, priors, and likelihood functions.
- Uncertainty Estimation - Calculates uncertainty bounds for algorithms and forecasts to determine the reliability of outcomes.
- Variational Inference - Provides variational inference tools to approximate complex posterior distributions through functional optimization.
- Bayesian Probabilistic Programming Frameworks - Serves as a complete Bayesian probabilistic programming framework for building models and performing inference.
- Computational Backend Integrations - Integrates with high-performance external backends to accelerate posterior estimation and complex tensor computations.
- Probability Distributions - Allows the specification of observed likelihoods and unobserved priors using a wide variety of probability distributions.
- Univariate Normal Distributions - Enables the creation of tensor variables representing normal distributions for statistical modeling.
- Probabilistic Programming Workflows - Supports a complete probabilistic programming workflow for designing generative models that integrate differential equations.
- Automatic Differentiation Frameworks - Provides a comprehensive automatic differentiation engine to compute gradients for inference algorithms and ODE solvers.
- Causal Inference Tools - Enables causal inference analysis to identify cause-effect relationships and simulate counterfactual scenarios.
- Computational Graphs - Represents probabilistic models as directed computational graphs to optimize the calculation of log-probabilities and gradients.
- Gaussian Processes - Models complex functions and spatial correlations using kernels and sparse approximations for regression.
- GPU-Accelerated Inference - Features a high-performance backend that compiles models into machine code for GPU-accelerated inference.
- Gradient Computation - Calculates precise gradients of joint distributions to enable efficient gradient-based sampling and optimization.
- Generalized Linear Models - Supports the construction of generalized linear models using Bayesian priors to estimate variable relationships.
- Mixture Model Estimation - Models complex distributions as a combination of multiple sub-distributions such as Gaussian or Dirichlet mixtures.
- Seismic Behavioral Models - Performs Bayesian analysis on seismic data to model behavioral patterns of earthquakes.
- Posterior Predictive Checks - Evaluates model fit by comparing observed data to simulated data from prior and posterior distributions.
- Model Data Updating - Swaps values or shapes of data containers within a symbolic model to perform predictions on new datasets.
- Posterior Analysis - Calculates summary statistics and generates diagnostic plots for exploratory analysis of Bayesian models.
- Predictive Sampling - Allows producing predictions on new data by applying sampled posterior distributions for predictive checks.
- Prior Predictive Sampling - Enables drawing samples from the prior predictive distribution to evaluate model assumptions before observing data.
- Counterfactual Analyses - The project simulates new data under hypothetical scenarios by modifying specific model parameters.
- Survival Analysis - Provides survival analysis capabilities to model time-to-event data using censored data and frailty regression.
- Tensor Computation Backends - Implements a backend-agnostic tensor execution system that leverages high-performance libraries for CPU and GPU computations.
- Time Series Forecasting - Implements time series forecasting using autoregressive models, stochastic volatility, and state-space frameworks.
- Probabilistic Forecasting - Provides probabilistic forecasting for time series data, generating uncertainty intervals rather than just point estimates.
- Convergence Monitoring - Provides tools to analyze divergences and sampler statistics to diagnose MCMC convergence.
- Bayesian Additive Regression Trees - Combines decision trees with Bayesian methods to perform categorical, quantile, and heteroscedastic regression.
- Exoplanetary Models - Analyzes transit and radial velocity observations of exoplanets and astronomical time series.
- A/B Testing - Analyzes experimental results to compare product versions and quantify the impact of changes using Bayesian methods.
- Just-In-Time Compilation - Utilizes just-in-time compilation to translate symbolic model definitions into optimized machine code for faster execution.
- Differential Equation Solvers - Integrates ordinary differential equations within probabilistic models to infer parameters from dynamic systems.
- Model Result Analysis - Generates diagnostic plots and computes summary statistics to evaluate the results of probabilistic models.
- Custom Distributions - Allows the definition of arbitrary probability distributions via custom log-probability functions.
- Artificial Intelligence - Library for Bayesian statistical modeling and probabilistic machine learning.
- Frameworks d'apprentissage automatique - A library for probabilistic programming and Bayesian statistical modeling.
- Probabilistic Modeling - Framework for Bayesian stochastic modeling and inference.
- Scientific Computing - Probabilistic programming and Bayesian modeling.
- Data Analysis Visualization - MCMC sampling toolkit for Python.
- Data Science - Probabilistic programming and Bayesian modeling.
- Data Science and Databases - Probabilistic programming and Bayesian modeling.
- Bibliothèques numériques - Probabilistic programming for Bayesian modeling and machine learning.
- Statistical Modeling - Toolkit for Markov Chain Monte Carlo sampling.
- Python Projects - Listed in the “Python Projects” section of the Awesome For Beginners awesome list.
- Scientific Computing Libraries - Library for Bayesian modeling and probabilistic machine learning.
- Statistical Modeling - Flexible Bayesian modeling and probabilistic programming.
- Scientific Computing - Listed in the “Scientific Computing” section of the Awesome Python awesome list.
