rust-ndarray/ndarray
Ndarray
ndarray 是一个 Rust 的多维数组库，用作线性代数框架和科学计算工具。它提供了创建和操作 n 维数组的核心基础设施，既充当并行数组处理器，也充当数值数据分析工具包。
该库的独特之处在于提供高效的切片和内存视图，允许在不复制的情况下共享数据。它利用优化的后端数学库进行高速矩阵乘法，并将繁重的数学迭代分布到多个 CPU 线程上以加速处理。
该项目涵盖了广泛的数学运算，包括逐元素算术、基于轴的数据聚合和点积计算。它还包括用于数组操作的全面实用程序，如重塑、展平、堆叠和坐标网格生成，以及对随机数组生成和序列化的支持。
Features
- Multidimensional Arrays - Provides the core infrastructure for managing and manipulating n-dimensional arrays and matrices.
- Array View Creation - Produces read-only or mutable windows into existing data buffers without copying underlying elements.
- Strided - Maps multidimensional indices to flat memory buffers using axis-specific step sizes for efficient zero-copy slicing.
- Zero-Copy Array Views - Provides lightweight references to existing memory buffers to allow slicing and reshaping without data duplication.
- Tensor Slicing and Indexing - Extracts specific elements or sub-regions of arrays using coordinate-based indexing and slicing.
- Numeric Type Traits - Uses generics and numeric traits to apply mathematical operations across different numeric types and dimensions.
- Array Layout Management - Modifies shape and stride information independently of the underlying data using dedicated reference types.
- Array Slicing - Extracts sub-sections of arrays using arbitrary step sizes and negative indices.
- Element-wise Array Operations - Executes high-performance element-wise operations and mathematical functions across multidimensional arrays.
- Generalized Matrix Multiplications - Executes high-performance matrix product computations with support for scalar scaling.
- High-Performance Array Arithmetic - Executes high-performance mathematical operations across arrays of varying dimensions using optimized routines.
- Scientific Computing - Provides a computational framework for performing complex mathematical modeling and multi-dimensional array operations.
- Linear Algebra Libraries - Serves as a high-performance framework for vector and matrix operations.
- Linear Algebra Routines - Implements fundamental linear algebra operations including matrix multiplication and dot products.
- Multi-Dimensional Arrays - Implements data structures that organize numeric elements into grids or higher-dimensional spaces.
- Multidimensional Array Containers - Offers dense data structures supporting various memory layouts for numerical computing and data analysis.
- N-Dimensional Array Libraries - Provides a comprehensive Rust library for creating and manipulating multidimensional arrays with efficient memory views.
- Linear Algebra - Provides high-performance routines for vector and matrix operations, including dot products.
- Parallel Array Processing - Increases processing speed by distributing array iterations and methods across multiple CPU threads.
- Dot Product Computation - Calculates the inner product of two arrays, including those with dynamic dimensions.
- Ownership and Borrowing - Uses Rust's ownership and borrowing semantics to separate data ownership from data access via unified interfaces.
- Multidimensional Shape Handling - Implements runtime-defined vectors to support arrays with an arbitrary number of axes.
- Dimension Squeezing - Removes dimensions of size one from an array to simplify its overall shape.
- Scientific Array Serialization - Supports persisting and loading multidimensional arrays using scientific data formats.
- Axis Permutation - Changes the order of dimensions or reverses the direction of specific axes in-place.
- Parallel Iterators - Distributes element-wise operations and axis reductions across multiple CPU cores using parallel iterators.
- BLAS Backend Integration - Leverages optimized backend math libraries for high-speed floating-point matrix multiplication.
- Strided Window Generation - Generates overlapping windows of a specific size along an axis using a defined step size.
- Ownership-Based Memory Management - Implements memory management using Rust's ownership and borrowing model to handle owned arrays and non-owning views.
- Array Function Mapping - Applies custom functions to corresponding elements of multiple arrays in lock step with broadcasting support.
- Array Combinations - Joins arrays together by stacking them along new axes or concatenating them along existing ones.
- Array Collapsing - Collapses multidimensional arrays into one-dimensional sequences while preserving element order.
- Array Reshaping - Changes the dimensions of an array while preserving elements and specifying memory layout order.
- Array Splitting - Divides single arrays into multiple smaller views along a specified axis.
- Dimensional Reductions - Reduces arrays by computing sums and other aggregations along specified dimensions.
- BLAS and LAPACK Integrations - Provides interfaces to optimized BLAS libraries for high-performance linear algebra and matrix multiplication.
- Coordinate Grid Generation - Builds regular grids of values to evaluate functions over a specific coordinate space.
- Numerical Computation Accelerations - Accelerates mathematical processing by distributing array operations across multiple CPU threads.
- Parallel Array Processors - Distributes heavy mathematical iterations and array methods across multiple CPU threads to accelerate performance.
- Scientific Computing Libraries - N-dimensional array library for Rust.
