What are the main features of pythonot/pot?

The main features of pythonot/pot are: Exact Solvers, Gromov-Wasserstein Distance Computations, Optimal Transport Libraries, Optimal Transport Solvers, Wasserstein Distance Estimators, Cross-Framework Tensor Dispatch, Differentiable Optimization Layers, Differentiable Transport Solvers.

POT | Awesome Repos

POT is an optimal transport library providing a collection of solvers for computing Wasserstein, Gromov-Wasserstein, and Fused Gromov-Wasserstein distances between probability distributions. It functions as a differentiable tensor framework that integrates with various tensor libraries to enable automatic differentiation and GPU acceleration.

The project is distinguished by its ability to align data distributions across different metric spaces by comparing internal relational structures rather than coordinates. It implements mathematical optimization algorithms as differentiable layers, allowing for gradient-based updates within neural network workflows.

The toolkit covers a broad range of capabilities, including domain adaptation for aligning source and target distributions, the computation of various barycenters for distributions and graphs, and the estimation of transport mappings. It also provides tools for graph analysis, such as subgraph matching and dictionary learning, as well as dimensionality reduction techniques that preserve Wasserstein distance structures.

Solvers are implemented through a backend-agnostic tensor interface to support high-performance hardware acceleration across multiple tensor frameworks.

Features

Exact Solvers - Provides exact solvers and specialized 1D solvers to calculate precise optimal transport plans.
Gromov-Wasserstein Distance Computations - Implements solvers for estimating distances and transport plans between distributions in different metric spaces based on relational geometry.
Optimal Transport Libraries - Provides a comprehensive collection of solvers for computing Wasserstein, Gromov-Wasserstein, and Fused Gromov-Wasserstein distances.
Optimal Transport Solvers - Provides a comprehensive suite of mathematical solvers for computing optimal transport matrices between discrete probability distributions.
Wasserstein Distance Estimators - Provides solvers to calculate the minimum cost of transforming one probability distribution into another.
Cross-Framework Tensor Dispatch - Provides an abstraction layer that allows tensor operations to run across multiple backends including PyTorch, TensorFlow, and JAX.
Differentiable Optimization Layers - Implements mathematical optimization algorithms as differentiable layers for gradient-based updates in neural networks.
Differentiable Transport Solvers - Implements optimal transport algorithms as differentiable layers for seamless integration into gradient-based machine learning workflows.
Sinkhorn Solvers - Provides fast approximation of transport plans using iterative Sinkhorn solvers.
Optimal Transport Layers - Integrates differentiable optimal transport solvers directly into graph neural network architectures as layers.
Hardware Acceleration Backends - Integrates with hardware acceleration backends to enable GPU-powered solver execution and automatic differentiation.
Tensor Computation Backends - Integrates with various tensor backends to enable automatic differentiation and high-performance hardware acceleration for transport operations.
Transport Solver Integration - Incorporates differentiable transport solvers directly into neural network workflows for gradient-based optimization.
Discrete Solvers - Solves optimal transport problems for discrete distributions using linear programming and entropic regularization.
Entropic Solvers - Uses entropy regularization and the Sinkhorn algorithm to approximate optimal transport plans.
Regularized Solvers - Ships transport solvers that incorporate regularization terms via conditional gradient methods for improved speed.
Unbalanced Transport Solvers - Computes optimal transport plans between distributions with different total masses using the Sinkhorn algorithm.
Wasserstein Barycenter Analysis - Computes representative average distributions, known as barycenters, based on weighted optimal transport distance.
Entropic Regularization - Accelerates optimal transport computations by adding an entropy term to the objective function.
Optimal Transport Solvers - Computes optimal transport matrices to minimize mass movement costs between discrete distributions.
Graph Topology Alignment - Matches structured graphs by measuring the distance between their internal geometries and associated node features.
Distributional Alignment - Implements mathematical utilities to align data distributions across different metric spaces to improve model generalization.
Distributional Domain Alignment - Aligns probability distributions across different domains using optimal transport to transport labels and improve model generalization.
GPU Acceleration Backends - Provides a backend-agnostic interface to execute complex optimal transport solvers on GPU hardware.
Tensor Backend Switching - Executes transport solvers across multiple tensor libraries like PyTorch and JAX through a common backend-agnostic interface.
Computational Backend Integrations - Implements interfaces to delegate heavy mathematical operations to high-performance external tensor libraries for acceleration.
Differentiable Tensor Backends - Functions as a differentiable tensor framework that integrates with multiple libraries for GPU acceleration.
Gromov-Wasserstein Metrics - Measures distance between distributions in different metric spaces by analyzing internal relational geometry.
Partial Gromov-Wasserstein Distances - Provides solvers for measuring distance between different metric spaces where only a portion of the mass is transported.
Partial Wasserstein Distances - Calculates optimal transport plans and distances when only a specified fraction of the probability mass is transported.
Distribution Alignment Toolkits - Provides a comprehensive suite of tools for mapping and aligning probability distributions using various transport mappings.
Fused Gromov-Wasserstein Distances - Implements metrics that combine Wasserstein and Gromov distances to align graphs using both topology and node features.
Graph Barycenters - Computes representative average structures between multiple graphs while preserving topology and attributes.
Fused Gromov-Wasserstein Distances - Combines structural relational costs with node feature similarities to measure distance between graphs.
Gromov-Wasserstein Graph Distances - Provides solvers to compute the Gromov-Wasserstein divergence to minimize distances between graph topologies.
Metric Space Alignments - Aligns probability distributions across different metric spaces by comparing internal distance matrices.
Relational - Compares relational structures and internal geometries of distributions rather than using absolute coordinates.
Optimal Transport Distances - Calculates the minimum cost required to transform one probability distribution into another.
Optimal Transport Plan Computations - Calculates optimal transport coupling matrices to determine the most efficient way to move mass between probability distributions.
Semi-Relaxed Optimal Transport - Provides transport solvers that allow node reweighting to minimize divergence between structured datasets.
Sliced Wasserstein Distances - Estimates high-dimensional distances by averaging the Wasserstein costs of multiple one-dimensional linear projections.
Transport Map Estimations - Computes functional transformations, or transport maps, that push one probability distribution onto another.
Wasserstein Barycenters - Computes representative central distributions between probability measures using linear programming or entropic regularization.
Gromov-Wasserstein Barycenters - Provides computations for barycenters based on the Gromov-Wasserstein distance for distributions in different metric spaces.
Unbalanced Barycenters - Computes regularized Wasserstein barycenters for probability distributions that do not share the same total mass.
Fused-Cost Objectives - Features cost objectives that combine structural relational data and feature distances for complex dataset alignment.
Sparse Transport Solvers - Includes algorithms designed to generate sparse transport plans through specific regularization techniques.
Dimensionality Reduction - Simplifies complex datasets by reducing dimensionality while preserving the underlying Wasserstein distance structure.
Wasserstein Discriminant Analysis - Uses linear projection techniques based on Wasserstein distances to maximize separation between class distributions.
Gaussian Transport Plans - Computes specific optimal transport plans and densities tailored for Gaussian Mixture Models.
GMM Distribution Mappings - Transforms data between Gaussian Mixture Models using barycentric projections or sampling based on an optimal coupling.
Graph Dictionary Learning - Learns representative atoms for graph-structured data capturing both node attributes and connectivity.
Multi-Source Domain Adaptation - Aligns multiple source data distributions to a single target domain using optimal transport maps.
Partial Transport Solvers - Finds optimal transport plans for scenarios where only a fixed portion of the total mass is moved.
Smooth Transport Solvers - Produces smooth transport plans using regularization techniques such as KL divergence or L2 norm.
Stochastic Transport Solvers - Provides solvers that utilize stochastic optimization to compute transportation matrices for probability measures.
Transport Regularizations - Applies entropic, quadratic, or group Lasso regularization to ensure uniqueness and sparsity in transport plans.
Labeled Graph Barycenters - Calculates average graphs for labeled data using Fused Gromov-Wasserstein distance to preserve features and structure.
Linear Transport Mappings - Estimates parametric linear mappings between two distributions to approximate optimal transport plans for data transformation.
High-Dimensional Distribution Analysis - Analyzes complex high-dimensional probability distributions using slicing and projection techniques to reduce complexity.
Domain Adaptation Techniques - Provides algorithmic techniques for adapting data distributions to reduce shift and improve model performance.
Brenier Potentials - Computes strongly convex potentials to approximate the optimal transport map between two distributions.
Co-Optimal Transport Alignments - Aligns two dimensions of a data matrix simultaneously by solving co-optimal transport problems.
Circular Wasserstein Distances - Calculates optimal transport distances between probability distributions while accounting for circular periodicity.
Partial - Combines structural and feature costs to match only a fraction of the total mass between distributions.
Graph Analysis Frameworks - Provides algorithms for subgraph matching, graph dictionary learning, and computing barycenters of labeled graphs.
Subgraph Matching - Identifies matching sub-structures within graphs by combining structural adjacency and node features.
Quantized Distance Computations - Approximates transport distances by partitioning distributions or graphs into representative clusters for efficiency.
Subgraph Matching - Identifies similar sub-structures between graphs by computing distances between adjacency matrices.
Low-Rank Coupling Approximations - Reduces computational complexity of large transport plans using factored low-rank coupling approximations.
Large-Scale Transport Solvers - Processes massive datasets using stochastic solvers and differentiable losses to maintain computational efficiency.
Monge Map Approximations - Provides numerical approximations of Monge maps that push one probability distribution onto another.
Monge Mapping Estimators - Approximates the optimal Monge map between distributions using Gaussian solutions, kernels, or barycentric mapping.
Marginal Constraint Relaxations - Handles unbalanced transport by replacing strict mass constraints with penalization terms.
Optimal Transport Cost Functions - Provides mathematical objectives to measure discrepancy by combining feature and structural costs in optimal transport.
Projective Distance Estimations - Estimates distances between high-dimensional distributions by averaging transport costs of 1D projections.
Spherical Sliced Discrepancies - Measures the discrepancy between probability distributions on a spherical domain using random projections.
Barycenter Debiasing - Implements post-processing and refinement to remove regularization bias from Sinkhorn barycenters.
Debiased Barycenters - Removes systematic bias introduced by entropic regularization to produce more accurate barycenter representations.
Stochastic Distance Estimators - Approximates distances between probability distributions using sampling and stochastic processes to improve scalability.
Entropic Dual Variables - Computes entropic dual variables used as potential functions to train neural networks via stochastic continuous distributions.
Lazy Tensor Evaluations - Employs lazy tensor evaluation to minimize memory overhead when processing large-scale mathematical computations.

lululxvi/deepxde

3,874在 GitHub 上查看

DeepXDE is a scientific machine learning library and deep learning PDE solver used to compute solutions for forward and inverse ordinary, partial, and integro-differential equations. It functions as a physics-informed neural network library that embeds physical laws and boundary conditions directly into the neural network loss function. The project provides a deep operator network framework for learning operator mappings that approximate relationships between functions in multiphysics problems. It is implemented as a multi-backend tensor library, allowing the system to switch between differen

pymc-devs/pymc

9,650在 GitHub 上查看

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 o

kornia/kornia

11,238在 GitHub 上查看

Kornia is a differentiable computer vision library and cross-framework tensor vision toolset. It implements vision operations as differentiable tensors to enable integration into deep learning pipelines and supports the transpilation of operations across PyTorch, TensorFlow, JAX, and NumPy. The project provides specialized toolsets for geometric vision and stereo depth, including algorithms for 3D scene reconstruction, camera calibration, and pose estimation. It further distinguishes itself as a differentiable image augmentation framework, applying random geometric and color transformations w

fchollet/keras

64,095在 GitHub 上查看

Keras is a high-level deep learning API used to design, build, and train neural networks for tasks such as computer vision, natural language processing, and time series forecasting. It provides a framework for defining model architectures and optimizing weights through a structured interface. The project is defined by a backend-agnostic design that allows the same model code to run across different compute engines. This multi-backend execution enables users to swap underlying engines to optimize for specific hardware or performance requirements. The system supports distributed model training

PythonOTPOT

POT

Features