This repo implements Dirichlet Process Mixture Modeling (DPMM) in WebGPU. DPMM is a hierarchical clustering model used to understand data clusters without knowing the number of clusters ahead of time. This means that DPMMs remain flexible and provide good uncertainty guarantees when the number of clusters is unknown.
Since this algorithm works locally, it works on any device without a backend (see here).