Torchcluster is a python package for cluster analysis.
Torchcluster is a python package for cluster analysis. The speed of the clustering algorithm has been effectively improved with the Pytorch backend. We are also working on test datasets and visualization tools. Related work is coming in the next release.
torchcluster should work on
- all Linux distributions no earlier than Ubuntu 16.04
- macOS X
- Windows 10
torchcluster also requires Python 3.5 or later. Python 2 support is coming.
Right now, torchcluster works on PyTorch 0.4.1.
pip install torchcluster
conda install -c tczhangzhi torchcluster
How torchcluster looks like
Define a dataset generator and generate a dataset:
from torchcluster.dataset.simple import SimpleDataset dataset_factory = SimpleDataset(2, feature=2, sigma=2, device=device) dataset = dataset_factory(100)
Configuring a clustering algorithm and get your result:
from torchcluster.zoo.spectrum import SpectrumClustering cluster = SpectrumClustering(2) result, _ = cluster(dataset)
You can also cluster your own data sets. The dataset should be a tensor of n by m, where n is the number of data points in the dataset and m is the dimension of each data point:
dataset = torch.cat([torch.randn(500,2) + torch.Tensor([-2,-3]), torch.randn(500,2) + torch.Tensor([2,1])])
Use spectral clustering to get the following results:
tensor([0, 0, ..., 1, 1])
Copyright (c) 2019-present, Zhang Zhi
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size torchcluster-0.1.4.tar.gz (4.7 kB)||File type Source||Python version None||Upload date||Hashes View|