NO， I’m referring to the result. you means there might be so many zero features in the als result ？
I think it is not related to the initial state, but I do not know why the percent of zero-vector is so high(50% around) > 在 2015年4月2日，下午6:08，Sean Owen <so...@cloudera.com> 写道： > > You're referring to the initialization, not the result, right? It's possible > that the resulting weight vectors are sparse although this looks surprising > to me. But it is not related to the initial state, right? > > On Thu, Apr 2, 2015 at 10:43 AM, lisendong <lisend...@163.com > <mailto:lisend...@163.com>> wrote: > I found that there are about 50% all-zero vectors in the ALS result ( both in > userFeatures and productFeatures) > > the result looks like this: > > <PastedGraphic-1.tiff> > > > I looked into the ALS.scala, the user and product factors seems to be > initialized by a gaussian distribution, so it should not be all-zero vector, > right? > > > 邮件带有附件预览链接，若您转发或回复此邮件时不希望对方预览附件，建议您手动删除链接。 > 共有 1 个附件 > PastedGraphic-1.tiff(607K) > 极速下载 > <http://preview.mail.163.com/xdownload?filename=PastedGraphic-1.tiff&mid=1tbiDQnPDVQG6QKGSwAAsy&part=3&sign=60d32e9f90d3dd36c5858328dd96eabe&time=1427969402&uid=lisendong%40163.com>
