tag:blogger.com,1999:blog-7332477090953898587.post3714130046800080337..comments2022-11-23T03:21:59.265+01:00Comments on IT Shared: Apache Mahout Samsara: The Quick Start IT Sharedhttp://www.blogger.com/profile/11763288513127086508noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-7332477090953898587.post-58359498895356738282015-12-29T08:35:37.305+01:002015-12-29T08:35:37.305+01:00very well organized and expalinedvery well organized and expalinedBhupendrahttps://www.blogger.com/profile/07894893642503968565noreply@blogger.comtag:blogger.com,1999:blog-7332477090953898587.post-84900448995904467182015-06-17T09:43:52.120+02:002015-06-17T09:43:52.120+02:00Thank you for comments, Dmitriy!
`drmX cbind 1` -...Thank you for comments, Dmitriy! <br />`drmX cbind 1` - this is really cool! Anonymoushttps://www.blogger.com/profile/03991393318576196455noreply@blogger.comtag:blogger.com,1999:blog-7332477090953898587.post-27625736180180836012015-06-16T22:12:26.168+02:002015-06-16T22:12:26.168+02:00Great write-up!
Also, DSL in 0.10.x branch is a l...Great write-up!<br /><br />Also, DSL in 0.10.x branch is a little more evolved now. E.g. vec * vec could be simply `vec ^= 2` or addBias routine is simply `drmX cbind 1` and there are also uniform samplers of rows as a part of standard api. (not perfect though. Can't contribute the perfect one at this point :). `mxX.plus(mxY)` is java style, should really be just `X + Y`.Dmitriy Lyubimovhttps://www.blogger.com/profile/02449945453685411163noreply@blogger.comtag:blogger.com,1999:blog-7332477090953898587.post-23465803668503922202015-06-16T22:05:00.496+02:002015-06-16T22:05:00.496+02:00Matrix recenter is problematic with sparse algebra...Matrix recenter is problematic with sparse algebra as it creates a dense matrix out of potentially fairly sparse matrix (and most problems for reduction are pretty sparse). Mahout Samsara has (d)spca method that avoids this density problem.Dmitriy Lyubimovhttps://www.blogger.com/profile/02449945453685411163noreply@blogger.com