This is no longer necessary and was in fact causing problems for all environments. http://www.benjaminfleischer.com/2013/12/05/testing-rubinius-on-travis-ci/
Problem with Hyperpipes was that in case of a tie, the last value in the sequence should win. See http://www.csee.wvu.edu/~timm/tmp/r7.pdf for a description. This involved creating a new Votes class (and accompanying tests) to tally votes and return the winner. The multilayer perceptron had a bug where the value of the index accumulator in data_to_input was being reassigned rather than incremented with the length of the current domain.
… change history at github, if needed.
…dex for 0.upto, the method was returning a fixnum instead of a enum