https://github.com/dnbaker/wmh

Weighted Minhash Code
https://github.com/dnbaker/wmh

Last synced: 20 days ago
JSON representation

Weighted Minhash Code

• Host: GitHub
• URL: https://github.com/dnbaker/wmh
• Owner: dnbaker
• Created: 2021-01-31T22:04:29.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2021-02-21T00:14:02.000Z (almost 4 years ago)
• Last Synced: 2023-03-01T16:46:27.146Z (almost 2 years ago)
• Language: C++
• Size: 35.2 KB
• Stars: 5
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
Contains re-implementations of data structures from https://dl.acm.org/doi/10.1145/3219819.3220089 and https://arxiv.org/abs/1911.00675, as well as Python bindings.

Installation:

```bash

git clone --recursive https://github.com/dnbaker/wmh

cd wmh/python

python3 setup.py install

```

Use:

```python

import wmh

import numpy as np

weights = np.random.rand(1000)

ids = np.random(100000, size=(1000,))

signature_size = 500

sigs = [wmh.hash(weights, ids, m=signature_size, stype=sketcher) for sketcher in

        ("bmh1", "bmh2", "pmh1", "phm1a")]

```

PMinHash effectively normalizes all counts, while bmh treats the id, w pairs as weighted elements in a weighted set.