curbside-clojure-ml

curbside-clojure-ml is a library used to train and optimize models, as well as to make predictions. It implements the following ML algorithms:

• lsvm
• svm
• c4
• random-forest
• m5p
• xgboost

Usage

From your application code, you will need to require the models interface:

(require '[curbside.ml.models])

The model namespace offers the following interface, based on the provided algorithm keyword (i.e. :xgboost)

• save
• load
• dispose
• train
• evaluate
• optimize-hyperparameters
• infer

Benchmarking

Use lein libra to run benchmarks. Add benchmarks under bench/.

Releasing to Github

We use https://github.com/relaynetwork/lein-release to release artifacts to Github. To use:

1. Export GITHUB_ACTOR and GITHUB_TOKEN. You can generate a token from your own account settings with the write:packages permission.
2. Make sure you are on the master branch, and that its remote branch is set to the primary source repository (https://github.com/Curbside/curbside-clojure-ml).
3. We use https://semver.org to describe releases, so make sure to update the version accordingly in project.clj
4. Release a new version with lein deploy release