[ https://issues.apache.org/jira/browse/SPARK-17027?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15418065#comment-15418065 ]
Sean Owen commented on SPARK-17027: ----------------------------------- Is the problem in the naive calculation of n choose k? {code} private def choose(n: Int, k: Int): Int = { Range(n, n - k, -1).product / Range(k, 1, -1).product } {code} Let's just call http://commons.apache.org/proper/commons-math/apidocs/org/apache/commons/math3/util/CombinatoricsUtils.html#binomialCoefficient(int,%20int) > PolynomialExpansion.choose is prone to integer overflow > -------------------------------------------------------- > > Key: SPARK-17027 > URL: https://issues.apache.org/jira/browse/SPARK-17027 > Project: Spark > Issue Type: Bug > Components: ML > Affects Versions: 1.6.0, 2.0.0 > Reporter: Maciej Szymkiewicz > Priority: Minor > > Current implementation computes power of k directly and because of that it is > susceptible to integer overflow on relatively small input (4 features, degree > equal 10). It would be better to use recursive formula instead. -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org