Great, many thanks for the contribution, will review till the end of this week
пн, 31 авг. 2020 г. в 09:05, Mark Andreev <mark.andr...@gmail.com>: > Hi, > > I have added test cases to jira and PR with these cases. > > Jira: https://issues.apache.org/jira/browse/IGNITE-13392 > PR: https://github.com/apache/ignite/pull/8196 > > Please review and share feedback. > > On Sun, 30 Aug 2020 at 12:48, Alexey Zinoviev <zaleslaw....@gmail.com> > wrote: > > > Great, could you provide in comment some test cases with correct from > > formula or wolfram and wrong answers > > > > сб, 29 авг. 2020 г., 22:25 Mark Andreev (Jira) <j...@apache.org>: > > > > > Mark Andreev created IGNITE-13392: > > > ------------------------------------- > > > > > > Summary: Incorrect Vector::kNorm evaluation for odd powers > > > Key: IGNITE-13392 > > > URL: > https://issues.apache.org/jira/browse/IGNITE-13392 > > > Project: Ignite > > > Issue Type: Improvement > > > Components: ml > > > Reporter: Mark Andreev > > > Assignee: Mark Andreev > > > > > > > > > Current implementation of `Vector::kNorm` is incorrect. > > > > > > Current formula is > > > (`org.apache.ignite.ml.math.primitives.vector.AbstractVector:882`): > > > {code:java} > > > (\sum_{i}{x^p})^{1/p} > > > {code} > > > But correct formula is: > > > {code:java} > > > (\sum_{i}{|x|^p})^{1/p} > > > {code} > > > We can verify this using lectures ([ > > > https://www.math.usm.edu/lambers/mat610/sum10/lecture2.pdf)] or using > > > Wolfram Mathematica: > > > {code:java} > > > > Norm[{x, y, z}, p] > > > (Abs[x]^p+Abs[y]^p+Abs[z]^p)^(1/p){code} > > > > > > > > > > > > > > > -- > > > This message was sent by Atlassian Jira > > > (v8.3.4#803005) > > > > > > > > -- > Best regards, > Mark Andreev >
