Easy To Use

Lathe brings an entirely different methodology to the Julia language. Types are created to adhere to the object-oriented programming paradigm, and syntax is akin to that of Pythonic machine-learning packages, like SkLearn.

Fully Featured

Lathe includes many of the tools commonly used by machine-learning engineers and scientists out of the box, rather than relying on more dependencies to do so.

Deployable

Lathe models can be easily serialized and deployed onto production servers using Genie.jl, or a similar high-performance web-server. Lathe also has support for pipelines, meaning most pre-processing operations can be automated and performed with one easy call.

Fast

Lathe uses a faster methodology than most other Julia packages for machine-learning. Furthermore, the package also takes advantage of the natural ability of the language to be fast. As a result, Lathe is also faster than most similar packages for other high-level statistical programming languages.

Julian

Lathe is written in 100-percent pure Julia. As a result, the package often takes advantage of very Julian methods of dealing with problems, such as dispatch, macros, and syntactical expressions.

Stats

• Distributions
• Statistical tests
• Bayesian tests
• Model validation
• Sampling
• General Statistics

Preprocess

• Scalers
• Encoders
• Splitters

Models

• Pipelines
• Powerlog
• Logistic Regression
• Kmeans Clustering
• Random Forest Classifier
• Decision Tree Classifier
• Pipelines
• Linear Regression
• Linear Least square

Documentation on JuliaHub

Example Notebooks (Lathe Books!)