Kotlin for data science
From building data pipelines to productionizing machine learning models, Kotlin can be a great choice for working with data:
Kotlin is concise, readable, and easy to learn.
Static typing and null safety help create reliable, maintainable code that is easy to troubleshoot.
Being a JVM language, Kotlin gives you great performance and an ability to leverage an entire ecosystem of tried and true Java libraries.
Notebooks such as Jupyter Notebook, Datalore, and Apache Zeppelin provide convenient tools for data visualization and exploratory research. Kotlin integrates with these tools to help you explore data, share your findings with colleagues, or build up your data science and machine learning skills.
Jupyter Kotlin kernel
The Jupyter Notebook is an open-source web application that allows you to create and share documents (aka "notebooks") that can contain code, visualizations, and Markdown text. Kotlin-jupyter is an open source project that brings Kotlin support to Jupyter Notebook.
Check out Kotlin kernel's GitHub repo for installation instructions, documentation, and examples.
Kotlin Notebooks in Datalore
With Datalore, you can use Kotlin in the browser straight out of the box, no installation required. You can also collaborate on Kotlin notebooks in real time, get smart coding assistance when writing code, and share results as interactive or static reports. Check out a sample report.
Zeppelin Kotlin interpreter
Apache Zeppelin is a popular web-based solution for interactive data analytics. It provides strong support for the Apache Spark cluster computing system, which is particularly useful for data engineering. Starting from version 0.9.0, Apache Zeppelin comes with bundled Kotlin interpreter.
The ecosystem of libraries for data-related tasks created by the Kotlin community is rapidly expanding. Here are some libraries that you may find useful:
Multik: multidimensional arrays in Kotlin. The library provides Kotlin-idiomatic, type- and dimension-safe API for mathematical operations over multidimensional arrays. Multik offers swappable JVM and native computational engines, and a combination of the two for optimal performance.
KotlinDL is a high-level Deep Learning API written in Kotlin and inspired by Keras. It offers simple APIs for training deep learning models from scratch, importing existing Keras models for inference, and leveraging transfer learning for tweaking existing pre-trained models to your tasks.
Kotlin DataFrame is a library for structured data processing. It aims to reconcile Kotlin's static typing with the dynamic nature of data by utilizing both the full power of the Kotlin language and the opportunities provided by intermittent code execution in Jupyter notebooks and REPLs.
Kotlin for Apache Spark adds a missing layer of compatibility between Kotlin and Apache Spark. It allows Kotlin developers to use familiar language features such as data classes, and lambda expressions as simple expressions in curly braces or method references.
kotlin-statistics is a library providing extension functions for exploratory and production statistics. It supports basic numeric list/sequence/array functions (from
skewness), slicing operators (such as
simpleRegressionBy), binning operations, discrete PDF sampling, naive bayes classifier, clustering, linear regression, and much more.
kmath is an experimental library that was intially inspired by NumPy but evolved to more flexible abstractions. It implements mathematical operations combined in algebraic structures over Kotlin types, defines APIs for linear structures, expressions, histograms, streaming operations, provides interchangeable wrappers over existing Java and Kotlin libraries including ND4J, Commons Math, Multik, and others.
krangl is a library inspired by R's dplyr and Python's pandas. This library provides functionality for data manipulation using a functional-style API; it also includes functions for filtering, transforming, aggregating, and reshaping tabular data.
lets-plot is a plotting library for statistical data written in Kotlin. Lets-Plot is multiplatform and can be used not only with JVM, but also with JS and Python.
londogard-nlp-toolkit is a library that provides utilities when working with natural language processing such as word/subword/sentence embeddings, word-frequencies, stopwords, stemming, and much more.
Since Kotlin provides first-class interop with Java, you can also use Java libraries for data science in your Kotlin code. Here are some examples of such libraries:
DeepLearning4J - a deep learning library for Java
ND4J - an efficient matrix math library for JVM
Dex - a Java-based data visualization tool
Smile - a comprehensive machine learning, natural language processing, linear algebra, graph, interpolation, and visualization system. Besides Java API, Smile also provides a functional Kotlin API along with Scala and Clojure API.
Smile-NLP-kt - a Kotlin rewrite of the Scala implicits for the natural language processing part of Smile in the format of extension functions and interfaces.
Apache Commons Math - a general math, statistics, and machine learning library for Java
NM Dev - a Java mathematical library that covers all of classical mathematics.
OptaPlanner - a solver utility for optimization planning problems
Charts - a scientific JavaFX charting library in development
Apache OpenNLP - a machine learning based toolkit for the processing of natural language text
CoreNLP - a natural language processing toolkit
Apache Mahout - a distributed framework for regression, clustering and recommendation
Weka - a collection of machine learning algorithms for data mining tasks
Tablesaw - a Java dataframe. It includes a visualization library based on Plot.ly
If this list doesn't cover your needs, you can find more options in the Kotlin Machine Learning Demos GitHub repository with showcases from Thomas Nield.