Some software I’ve written:
- salmon, a library for asking the next “best” triplet query to crowdsourcing participants.
- caption-contest-data, which aggregates human responses from The New Yorker’s Caption Contest. Human responses are collected with the voting interface.
- adadamp, an implementation of a method that adapts batch size to model quality. This method improves the convergence of SGD.
- swix, the Swift Matrix Library (archived).
- jem-press, a static site generator inspired by jem-doc with Markdown and Mathjax.
I also wrote some code and documentation for Dask-ML’s hyper-parameter optimization:
- Dask-ML’s page on “Hyper-parameter searches”. What problems arise in machine learning’s hyperparameter searches, and what tools can get around those problems?
Other software I’ve written without a dedicated documentation page: