from Hacker News

Experiments in Handwriting with a Neural Network

by swannodette on 12/6/16, 7:58 PM with 15 comments

• by ludwigschubert on 12/7/16, 3:03 AM

  This is incredible important research. There are fields in which adoption of ML techniques is dependent on it being explainable and inspectable to stakeholders, e.g. in certain health care, policy or finance applications.

  There is work being done on explaining non-linear models, such as LIME [1], but much more is needed. At a recent d3 meetup the topic was discussed but we primarily noted its importance. Notes at [2].

  [1] https://github.com/marcotcr/lime [2] https://docs.google.com/document/d/1o-DO65PEZShLuadyRd35PsTJ...

• by shancarter on 12/6/16, 8:12 PM

  Me and the other authors of the post will be around if anyone has any questions.

• by ludwigschubert on 12/7/16, 3:09 AM

  In "Examining the Internals of the Model" it could be interesting to show an overview activation map that sums over the absolute activation of all the examples—to highlight which cells activate on all examples—and one that is colored by variance of activation across examples—to show which cells influence the character/differences in writing.

• by nl on 12/6/16, 10:03 PM

  Nice! Is it trained on Latin character-based hand writing only, or can it also predict other charactersets?

• by Qwertystop on 12/6/16, 9:09 PM

  Quite interesting to see the inspection of the model.

• by devoply on 12/7/16, 12:53 AM

  We're slowly coming to an era where everything can be easily forged.