Stable distribution and data sketching
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Zoubin Ghahramani.
We introduce the stable law and its properties, and present an interesting application in dimension reduction over data streams known as data sketching. Of particular interest is the situation where the data stream is prohibitively large that it cannot be stored in main computer memory or on disk for subsequent access, but instead must be processed on the fly. The method of data sketching constructs a lower dimensional representation such that l_{alpha} norms and quasinorms of the data stream, for 0 < alpha <= 2, can be accurately recovered with significantly reduced computational cost. In the extreme case as alpha tends to 0, the norm converges to the Hamming distance, giving the cardinality of the data stream, while for alpha >= 1, the norm of the difference between two streams is a meaningful measure of dissimilarity.
This talk is part of the Machine Learning @ CUED series.
This talk is included in these lists:
Note that exdirectory lists are not shown.
