The aim was to make an easy to access machine learning library for Scala similar to scikit.

To Run

Run Main class in java/numscala/ui/Main.java

There are two sections to numscala. One is written in Scala and the other in Java for demo purposes.

• Scala

  • Activation Functions - Contains sigmoid, softplus and tanh function.
  • Linear Models - Houses Linear and Logistic Regression.
  • NonLinear Models - Houses Perceptron and Neural Network.
  • Test - Houses all the datasets, their respective loaders and the Demo classes for each model.
  • Utilities - Contains one file which consists of functions like Accuracy, Log Loss and Splitting the Training and Testing set.

• Java

  • Scala Intepreter - Contains service to convert Seq and Object from invoking Scala functions to Java's List and Double
  • MenuUI - Handles selection of model to train
  • NonGraphableUI - Handles GUI for Linear Regression, this model can't produce a graph
  • NonLinearModelUI - Handles GUI for Logistic Regression, Perceptron and Neural Network
  • View mode - Allows for more ease in observing a graph
  • Hover Mode - Allows to navigate the graph at the cost of it looking funky (The number of data points is reduced by a factor of 2 to counteract this problem)