Packages
Riskfolio-Lib builds optimized investment portfolios using mathematical models—mean-variance, risk parity, hierarchical clustering, and Black-Litterman approaches—with support for 26+ risk measures and constraints on leverage, turnover, and cardinality.
Polars OLS provides Rust-optimized linear regression models (OLS, WLS, Ridge, Elastic Net, non-negative least squares, recursive least squares) as Polars expressions, enabling efficient least-squares estimation within Polars workflows.
ProxSuite is a numerically robust quadratic programming solver that handles dense, sparse, and matrix-free problems with specialized algorithms for robotics and general optimization.
fpylll is a Python wrapper for the fplll lattice reduction library, providing access to algorithms like LLL and BKZ for reducing integer matrices and computing Gram-Schmidt orthogonalization.
However, the GPL version 2 or later license requires verification against your project's licensing constraints, and the unclear license treatment in metadata warrants…
Provides JAX-compatible implementations of elliptic integrals (ellipk, ellipkm1, ellipe) that are differentiable and JIT-compilable, addressing the gap in jax.scipy.special.
Install only if you actually use elliptic integrals; otherwise it adds an unnecessary dependency.
Netron is a viewer for neural network and machine learning models that supports ONNX, TensorFlow Lite, PyTorch, TensorFlow, Core ML, OpenVINO, Keras, Caffe, Darknet, Safetensors, NumPy, and several other formats.
Install it if you work with neural networks or machine learning models in any of the supported formats.
Dopamine is a research framework for prototyping reinforcement learning algorithms, providing implementations of DQN, C51, Rainbow, IQN, SAC, and PPO agents built primarily with JAX.
Enables ONNX Runtime to accelerate machine learning model inference on Intel hardware (CPUs, integrated/discrete GPUs, and NPUs) using OpenVINO optimizations.
minorminer finds graph minor embeddings—it maps one graph into another as a subgraph—using a configurable heuristic algorithm, plus specialized functions for clique and biclique embeddings on quantum-graph topologies.
Converts between two's complement integers and gray code (reflected binary code), where successive values differ by only one bit.
However, the abandoned maintenance status means no bug fixes or Python version updates are forthcoming—install only if the current functionality meets your needs and…
Provides Sage categories, basic rings, polynomials, and mathematical functions as a modularized pip-installable component of the passagemath mathematical software system.
However, review the GPL-2.0-or-later copyleft license for your use case, and note that Windows support is incomplete without WSL.
CyLP provides a Python interface to COIN-OR's linear and mixed-integer programming solvers (CLP, CBC, CGL), allowing you to model optimization problems and customize the solution process with Python-defined cut generators, branch-and-bound strategies, and pivot rules.
jaxlie implements Lie groups (SO2, SE2, SO3, SE3) for rigid body transformations in JAX, providing differentiable operations like exp, log, and matrix conversions for computer vision and robotics applications.
Cleanlab automatically detects and helps fix data quality issues—mislabeled examples, outliers, duplicates, and annotation errors—in datasets for machine learning, working with any model and any data modality.
Formulates nonlinear optimization models symbolically using NumPy-inspired syntax for use with D-Wave's Stride hybrid solver.
OTT-JAX solves optimal transport problems—matching and comparing point clouds—using JAX-accelerated algorithms like Sinkhorn, including support for Gromov-Wasserstein and Wasserstein barycenter computations.
Implements unit of measurement arithmetic with static type checking and protobuffer serialization support, allowing operations on quantities with compatible units.
skfolio is a Python library for portfolio optimization and risk management that integrates with scikit-learn to build, tune, cross-validate, and stress-test portfolio models using a unified interface.
Install it if you need portfolio optimization; skip it if you only need basic mean-variance calculations or prefer a different API.
Jina is a framework for building and deploying AI services that communicate via gRPC, HTTP, and WebSockets, with built-in support for scaling, containerization, and cloud deployment.
Not recommended if you prefer minimal dependencies or need cutting-edge feature velocity.
Implements the Ramer-Douglas-Peucker algorithm to reduce the number of points in a 2D or 3D curve while preserving its shape, with a simple Python interface and optional NumPy support.
Implements regression discontinuity (RD) design estimation, inference, and plotting using local polynomial methods with robust bias-corrected confidence intervals.
TensorFlow Data Validation (TFDV) computes summary statistics, detects anomalies, and generates data schemas for machine learning datasets at scale using Apache Beam and TensorFlow.
Provides 2D and 3D vector, matrix, quaternion, and geometry math operations for Python, updated for Python 3 compatibility.
Install only if you are maintaining legacy code that already depends on it.
Enables PyTorch to use DirectML for GPU acceleration on DirectX 12-compatible hardware from AMD, Intel, NVIDIA, and Qualcomm.
However, operator coverage is still developing—verify that your specific PyTorch operations are supported via the operator roadmap before committing to production use.
Builds confidence intervals from data samples using bootstrap resampling, with support for A/B testing and statistical power analysis.
PyCUDA provides Pythonic bindings to Nvidia's CUDA API, letting you write GPU-accelerated code in Python with automatic memory management, error handling, and access to CUDA's full driver API.
Python bindings for FLINT and Arb that provide exact integer and rational arithmetic, arbitrary-precision real and complex numbers with error tracking, and polynomial, power series, and matrix operations over these types.
Provides high-level APIs for training, evaluating, and exporting machine learning models, encapsulating the full model lifecycle.
Provides differentiable GPU-capable solvers for controlled differential equations (CDEs), enabling Neural Controlled Differential Equation models for irregular time series with support for backpropagation and memory-efficient adjoint methods.
However, proceed with caution: the project is aging (last release October 2021, no recent commits), so compatibility with the latest PyTorch versions and long-term…
Computes finite difference numerical derivatives and solves partial differential equations on arrays of any dimension, with support for arbitrary accuracy orders, boundary conditions, and compact (implicit) schemes.
pygmsh provides a Python interface to Gmsh, a mesh generation tool, allowing you to define 2D and 3D geometries and generate unstructured meshes from them.
However, maintenance is dormant—the last release was 2022-01-28.
Optlang formulates and solves linear, mixed-integer, and quadratic optimization problems using symbolic expressions from sympy, with pluggable solver backends including GLPK, CPLEX, and Gurobi.
Install it if you're doing operations research, scientific computing, or constraint-based modeling in Python.
Provides a Python interface to IBM CPLEX's optimization solver, enabling linear, quadratic, and mixed-integer programming from Python code.
However, verify that the Community Edition's restrictions (key limits, possible performance caps, commercial use terms) align with your project before committing to…
Provides a Python interface to libhomfly for computing Homfly polynomials of knots and links.
Generates OpenSimplex noise across 2D, 3D, and 4D coordinate spaces, with both scalar and NumPy array interfaces for performance-critical workloads.
However, maintenance is dormant (last release 835 days ago), so expect no active development or rapid bug fixes—verify that the current API and behavior suit your use…
Ripser computes persistent homology and persistence diagrams for point clouds and sparse/dense datasets, enabling topological data analysis through an interface wrapping the fast C++ Ripser engine.
PyFixest performs fast high-dimensional fixed effects regression in Python, supporting OLS, WLS, IV, GLMs, quantile regression, and difference-in-differences estimation with multiple inference and post-estimation tools.
Provides runtime-checkable protocols and type hints for working with numeric types beyond Python's built-in int and float, enabling type-safe interfaces that accept members of the numeric tower and other numeric primitives.
PyVCG generates verification conditions for SMT solvers like CVC5 and SMTLIB2, supporting a range of logical sorts, datatypes, and arithmetic operations to build and solve formal verification problems.
Modeling library for building and solving mathematical optimization problems using IBM CPLEX and CP Optimizer solvers, with two namespaces for linear/quadratic and constraint programming respectively.