distillpub / post--handwriting

Four Experiments in Handwriting with a Neural Network

post--handwriting

Multithreaded make

make -j 4

Command to build the pca.json

  cat assets/validation/cells/0.json assets/validation/cells/1.json assets/validation/cells/2.json assets/validation/cells/3.json assets/validation/cells/4.json assets/validation/cells/5.json assets/validation/cells/6.json assets/validation/cells/7.json assets/validation/cells/8.json | bin/pca > assets/pca.json

Command to run tsne for sorting (copy the output into metadata.json's sortedOrder)

cat assets/validation/cells/0.json | bin/sort

Licensing

The content of this Distill article is licensed as CC-BY 4.0. Code is released under the Apache 2.0 License, even if enclosed inside a notebook or served as part of the interactive article.

post--handwriting's Issues

The "Variation" parameter does not actually correspond to Boltzmann temperature

According the source code (https://github.com/distillpub/post--handwriting/blob/master/public/assets/model/model.js#L137) the variation parameter actually corresponds to applying to the temperature to the mixture probabilities, and multiplying all of the sigma_x and sigma_ys by the temperature. But this is not the same as adjusting a mixture of Gaussians by the temperature. This can easily be seen by considering a mixture of a single 1d Gaussian. The standard deviation of the updated distribution will be sigma * sqrt(T), not sigma * T.

"divine some meaning"

Third paragraph, last sentence reads better with "derive."

Typo in Footnote 3

Equivalent is misspelt as equivelant