In both k-means and Apriori, we design bitmaps to encode information n on the GPU and to facilitate fast counting. In k-means,the bitmap represents the membership of data objects to clusters;
in Apriori, the bitmap stores occurrences of items in transactions.Such data structures utilize the GPU efficiently and significantly speed up the computation.As a result, our k-means is up to five
times faster than the prior GPU-based implementation [10], and our Apriori is an order of magnitude faster than the best CPU-basedApriori [8] implementation.