I have a learned model in the form of an Eigen (http://eigen.tuxfamily.org/index.php?title=Main_Page) Matrix.

Size of my model is very important. I am exploring lossy and lossless compression methods. To do this, I am putting the Eigen Matrix to a basic array of float and attempting to compress this as a start. I am wondering if anyone has any recommendations?

Sample file that contains raw array of float - https://drive.google.com/file/d/0BzRRuz ... sp=sharing

original size - 232kb

"standard" compression -
- gzip - 219kb
- zlib using most aggressive - 219kb

compression targeting floating point -
zfp - loss-less mode gives barely any reduction. I am still attempting to verify lossy mode impact on the model...