What is this?

This is a tool to run some ML subroutines from the command line. I'm primarly using this as a learning experience for C++ and how to optimize the implementation of ML algorithms.

I'm also curious about certain ways to parallelize training machine learning algorithms, and hopefully developing them in a systems-heavy language like C++ will allow me to explore this.

Usage

The command line interface consists of three main commands.

Training

The first is train, which has the following usage. ./bin/ml-cli train <data> <labels> <outfile> [<opt-type>] data should be a filename for CSV containing only the numerical data to be used, and labels should be a filename for a column vector of the correct labels for each training example. outfile is the file that the model will be stored in, so that it can be used for training. opt-type describes the optimizer used for training, and can be any of the following:

• adam - Adam (default)
• sgd - Stochastic Gradient Descent
• minibatch - Minibatch Gradient Descent
• batch - Batch Gradient Descent

Fine tuning of the parameters for each of the optimizers is supported in the underlying implementation, but not yet available in the CLI.

Predicting

To predict new data usign an existing model, you should use the predict command, which has the following usage. ./bin/ml-cli predict <model file> <data> <outfile> model file is the path to the model outputted by the training step, and must be one produced by this program. data should be the path to a CSV file of data to predict. The command will produce output in the file at path outfile, where it will be represented as a column vector.

Measuring Accuracy

To compute the accuracy of predicted labels vs. actual labels, use the accuracy command, which computes the 0-1 loss between the two provided label vectors. It has the following usage. ./bin/ml-cli accuracy <predicted> <actual> predicted and actual should both be in column vector format, the same format produces by prediction.

See Available Commands

If you ever forget a command, just run ./bin/ml-cli help and it'll descibe all of the available commands.

Next To-Dos

• Neural networks
• AdaBoost

Instructions to Run on Ubuntu

• Install the Eigen libary for linear algebra into the lib directory
• Run make in the root directory, and then call ml-cli