Show HN: Mini-batch k-means clustering for PyTorch tensors
Category: ai-ml
Tags: pytorch, clustering, unsupervised-learning, gpu, online-learning
Score: 5.3/10 (Innovation: 3, Technical: 5, Documentation: 7, Utility: 6)
minibatch-kmeans implements Sculley's mini-batch k-means algorithm for PyTorch tensors, enabling GPU-accelerated clustering and online/streaming training via partial_fit. It fills a practical gap since scikit-learn's MiniBatchKMeans is CPU-only, making it useful for PyTorch users working with large or GPU-resident datasets.
Target audience: data scientists and ML engineers
Repository: https://github.com/mb-14/minibatch-kmeans · Python · MIT
View on Hacker News