ferray-stats 0.3.5

Statistical functions, reductions, sorting, and histograms for ferray
Documentation

ferray-stats

Statistical functions, reductions, sorting, histograms, and set operations for the ferray scientific computing library.

What's in this crate

  • Reductions: sum, mean, var, std, min, max, argmin, argmax with SIMD pairwise summation
  • NaN-aware: nansum, nanmean, nanvar, nanstd, nanmin, nanmax, nanargmin, nanargmax
  • Sorting: sort, argsort, partition, argpartition, sort_complex
  • Histograms: histogram, histogram2d, histogramdd, histogram_bin_edges, bincount, digitize, searchsorted
  • Set operations: unique, unique_values, unique_counts, unique_inverse, unique_all, intersect1d, union1d, setdiff1d, setxor1d
  • Correlation: corrcoef, cov, correlate
  • Other: ptp (peak-to-peak), average (weighted mean)
  • Axis-aware reductions with Rayon parallelism for large arrays

Performance

ferray's pairwise-sum implementation uses a 4096-element base chunk with 4-way SIMD accumulators, cutting horizontal-reduction overhead 16× vs the previous 256-element base. Result on a Raptor Lake 14700K vs NumPy 2.4.4:

Operation Size Speedup vs NumPy
var 1K 28.8× faster
std 1K 17.8× faster
mean 1K 13.4× faster
var 100K 2.6× faster
std 100K 2.7× faster
var 1M 3.4× faster
std 1M 3.1× faster
sum 1K 1.9× faster
sum 100K 1.5× faster
sum 1M 1.1× faster (DRAM-bandwidth ceiling)

Usage

use ferray_stats::{mean, std_, var};
use ferray_core::prelude::*;

let a = Array1::<f64>::linspace(0.0, 1.0, 1000)?;
let m = mean(&a, None)?;
let s = std_(&a, None, None)?;
let v = var(&a, None, None)?;

This crate is re-exported through the main ferray crate.

License

MIT OR Apache-2.0