Updates 2019/07/10

It is not about quantity in this update, but more about the importance of the changes.

After several weeks of incremental development and testing, I have pushed out changes regarding random split and cross-validation fold generation. To avoid duplication of code, I implemented a general framework for such operations, which is now utilized across the board (which also fixes some bugs).

Fixes

  • The transformers BaseName and DirName now handle placeholder abbreviations ("${...}") correctly.
  • adams-weka and adams-weka-lts: by moving to the generalized framework for random split and cross-validation fold generation, a number of bugs in aligning additional attributes to predictions have been fixed.

Changes

  • upgraded jshell-scripting to 0.1.0.
  • upgraded pdfbox to 2.0.16 to address https://nvd.nist.gov/vuln/detail/CVE-2019-0228
  • adams-ml: the file-based dataset preparation schemes for train test split and cross-validation now make use of the new generalized framework for such operations.
  • adams-weka and adams-weka-lts: the random split and cross-validation fold generators (both default and grouped) now make use of the new generalized framework for such operations.

Additions

  • adams-ml: added a generalized framework for random split and cross-validation fold generation, to be utilized in other data domains. Makes use of the work implemented for binning algorithms.