The second version of the DRESS Kit has been rewritten almost entirely from scratch based on user experience and feedback. Many of the core statistical methods have been revised to provide a more streamlined codebase, and the modular system has been reorganized more intuitively and logically. Most importantly, all of the machine learning model algorithms have been reimplemented to support new features, such as batch training and new hyperparameters, and to provide a significantly enhanced performance.
The first version of the DRESS Kit was built on a collection of JavaScript files that were developed for several loosely related academic research projects. Over time, the scripts were successively refactored and reorganized into a more coherent library to maximize reusability. Ultimately, the decision was made to release the source code on GitHub to promote academic collaboration and engagement.
- New features:
DRESS.oneHot
encoding. - Rename and improve the performance of multilayer perceptron.
- Performance improvements to decision tree algorithms.
- Change
DRESS.crossValidate
function signature. - Feature improvement to
DRESS.kNN
. - Various bug fixes and performance improvements.
- Change
DRESS.meanMode
function signature. - Fix a bug in the chi-squared distribution function.
- Performance improvements to
DRESS.linear
.
- New features:
DRESS.histograms
,DRESS.heatmaps
. - Various performance improvements.
- Initial release