walk 0.1.3

Compatibility shim: re-exports `graphops` (use `graphops` directly).
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40

# walk

[![crates.io](https://img.shields.io/crates/v/walk.svg)](https://crates.io/crates/walk)
[![Documentation](https://docs.rs/walk/badge.svg)](https://docs.rs/walk)
[![CI](https://github.com/arclabs561/walk/actions/workflows/ci.yml/badge.svg)](https://github.com/arclabs561/walk/actions/workflows/ci.yml)

Graph random-walk primitives:

- Unbiased random walks
- Node2Vec / Node2Vec-style biased walks (including precomputed alias tables)
- PageRank and personalized PageRank

## Determinism

Most APIs take a `WalkConfig { seed, .. }`. For fixed `seed` and identical inputs,
walk generation is intended to be reproducible.

## Graph interfaces

- `Graph`: returns owned neighbor lists (`Vec<usize>`) per query (simple but allocates per step)
- `GraphRef`: returns borrowed neighbor slices (`&[usize]`) per query (faster for walking)

## Sampling integration

This crate uses reservoir sampling via `sample_start_nodes_reservoir`, which is useful when
`node_count` is too large to materialize `0..node_count` just to choose a subset of starts.

## Examples

- `cargo run --example ppr_hard_pool`: generate a seeded SBM graph, compute PPR from an anchor,
  then sample a stochastic top-\(k\) candidate pool using Gumbel-top-k on \(\log(\mathrm{PPR})\).
  - uses `testdata/karate_club.edgelist` by default (small real graph)
  - set `WALK_DATASET=lesmis` or `WALK_DATASET=florentine` for other bundled graphs
  - set `WALK_EDGELIST=/path/to/edges.txt` to run on your own graph

## References (starting points)

- Page et al. (1999): PageRank.
- Grover & Leskovec (2016): Node2Vec.
- Walker (1974) / Vose (1991): alias method for O(1) categorical sampling.