K-means (and K-means++) involve a lot of synchronous tasks which makes it difficult to achieve task parallelism. However, data parallelism may be achieved to improve the performance of certain areas in the process. MapReduce technique can be applied as a starting point in the Assignment step of the algorithm.
For the time being, this could be implemented using native std::thread and WebAssembly support can be thought of a little later once any improvements using threads are observed.
K-means (and K-means++) involve a lot of synchronous tasks which makes it difficult to achieve task parallelism. However, data parallelism may be achieved to improve the performance of certain areas in the process. MapReduce technique can be applied as a starting point in the Assignment step of the algorithm.
For the time being, this could be implemented using native
std::threadand WebAssembly support can be thought of a little later once any improvements using threads are observed.