# SciPy — Fundamental Algorithms for Scientific Computing

> SciPy provides efficient numerical routines for optimization, integration, interpolation, linear algebra, signal processing, and statistics, all built on NumPy arrays.

## Install

Save as a script file and run:

# SciPy — Fundamental Algorithms for Scientific Computing

## Quick Use
```bash
pip install scipy
python -c "
from scipy import optimize
result = optimize.minimize(lambda x: (x-3)**2, x0=0)
print(f'Minimum at x={result.x[0]:.2f}')
"
```

## Introduction
SciPy is a core library in the Python scientific computing stack. It builds on NumPy to provide a collection of mathematical algorithms and convenience functions for optimization, integration, interpolation, eigenvalue problems, statistics, signal and image processing, and sparse matrix operations.

## What SciPy Does
- Provides numerical optimization solvers (minimize, least_squares, linear_programming)
- Implements numerical integration and ODE/PDE solvers
- Offers sparse matrix representations and solvers for large-scale linear algebra
- Includes signal processing tools (FFT, filtering, wavelets, spectral analysis)
- Bundles statistical distributions, hypothesis tests, and descriptive statistics

## Architecture Overview
SciPy is organized into subpackages, each focused on a domain: `scipy.optimize`, `scipy.linalg`, `scipy.sparse`, `scipy.signal`, `scipy.stats`, and others. Under the hood, most routines wrap battle-tested Fortran and C libraries (LAPACK, BLAS, FITPACK, QUADPACK, ODEPACK) via Cython or f2py bindings, giving Python access to decades of validated numerical code with near-native performance.

## Self-Hosting & Configuration
- Install via `pip install scipy` or `conda install scipy` for optimized BLAS/LAPACK
- Requires NumPy as a dependency; compatible with NumPy 1.x and 2.x
- Build from source with `python -m pip install .` using Meson build system
- Use `scipy.show_config()` to verify linked BLAS/LAPACK implementations
- Pin versions in requirements.txt for reproducible scientific workflows

## Key Features
- Over 20 optimization algorithms including global optimizers (differential_evolution, dual_annealing)
- Sparse matrix support in CSR, CSC, COO, and other formats with efficient solvers
- Spatial algorithms: KD-trees, Delaunay triangulation, convex hulls, Voronoi diagrams
- Interpolation with splines (univariate, bivariate, N-dimensional)
- Statistical distributions (over 100 continuous and discrete) with fitting and testing

## Comparison with Similar Tools
- **NumPy** — provides arrays and basic linear algebra; SciPy adds advanced algorithms on top of NumPy
- **MATLAB** — proprietary with similar capabilities; SciPy is free and integrates with the Python ecosystem
- **Statsmodels** — deeper statistical modeling (regression, time series); SciPy covers broader numerical methods
- **SymPy** — symbolic math; SciPy is purely numerical for fast floating-point computation
- **scikit-learn** — machine learning algorithms; SciPy provides the underlying numerical building blocks

## FAQ
**Q: What is the difference between NumPy and SciPy?**
A: NumPy provides the array data structure and basic operations. SciPy builds on it with specialized algorithms for optimization, integration, statistics, and more.

**Q: Does SciPy support GPU acceleration?**
A: SciPy is CPU-only. For GPU-accelerated equivalents, consider CuPy which mirrors the SciPy API on NVIDIA GPUs.

**Q: How do I choose an optimization algorithm?**
A: Use `minimize` with method='L-BFGS-B' for smooth problems with bounds, 'Nelder-Mead' for derivative-free optimization, or 'trust-constr' for constrained problems.

**Q: Is SciPy suitable for production use?**
A: Yes. SciPy wraps proven numerical libraries and is used in production across scientific computing, engineering, and finance.

## Sources
- https://github.com/scipy/scipy
- https://scipy.org/

---
Source: https://tokrepo.com/en/workflows/c0268701-40a1-11f1-9bc6-00163e2b0d79
Author: Script Depot