基于GPU加速+Pytorch的K-Means聚类实现-附项目源码-优质项目实战.zip

# Balanced K-Means clustering in PyTorch Balanced K-Means clustering in Pytorch with strong GPU acceleration. # Installation As easy as: `pip install balanced_kmeans` # Getting started First things first: Classical kmeans algorithm as easy as ``` from balanced_kmeans import kmeans # experiment constants N = 10000 batch_size = 10 num_clusters = 100 device = 'cuda' cluster_size = N // num_clusters X = torch.rand(batch_size, N, dim, device=device) choices, centers = kmeans(X, num_clusters=num_clusters) ``` Now, if you want balanced kmeans you can run: ``` from balanced_kmeans import kmeans_equal N = 10000 batch_size = 10 num_clusters = 100 device = 'cuda' cluster_size = N // num_clusters X = torch.rand(batch_size, N, dim, device=device) choices, centers = kmeans_equal(X, num_clusters=num_clusters) ``` By default, forge initialization scheme is used for initial cluster centers. However, you may change the initial cluster centers by providing the keyword argument `initial_state` to either `kmeans` or `kmeans_equal`.