This research explores advanced compression algorithms for center-based clustering, presenting the first nearly-linear time algorithm for k-median and k-means. It establishes an optimal coreset size and assesses its practical applications through empirical comparisons with various sampling strategies. Results indicate that Fast-Coreset achieves superior compression guarantees, yet uniform sampling can be adequate for reliable outcomes in well-behaved datasets. The study also highlights the importance of hybrid techniques balancing efficiency and accuracy while pointing out remaining open questions in the field of data compression and clustering. Ultimately, it signifies a notable progress in clustering performance optimization.
Our research presents the first nearly-linear time coreset algorithm for k-median and k-means, exhibiting optimal compression guarantees while retaining efficiency.
The experimental analysis reveals that while Fast-Coreset outperforms its rivals, uniform sampling suffices for effective compression in well-structured datasets.
Collection
[
|
...
]