📚 ReductionML

Disclaimer: ReductionML is my own personal project and experiment. ReductionML is an experiment and very much a work in progress. It is an opportunity for me to explore my own design ideas. Things are very rough and I am sure will change a lot.

ReductionML is a machine learning framework with solutions to a range of problems. It revolves around the concept of simplifying problems by breaking them down into more manageable components that already have solutions. This process is done by reductions as they reduce one problem to another. This approach draws inspiration from the VowpalWabbit, a project I hold in high regard and deeply value. In fact, if you are familiar with VowpalWabbit then you should be able to pick up ReductionML with ease.

Crates

Crate
reductionml-core
reductionml-cli

Getting started

Install the reml tool:

cargo install reductionml-cli

There is built in support for CLI completions. To enable them, run the following commands:

Bash (optional)

reml gen-completions bash > /usr/share/bash-completion/completions/reml

Fish (optional)

reml gen-completions fish > ~/.config/fish/completions/reml.fish

First steps

The following will fetch a small 100 example dataset in VW text format. It will then create a new configuration file for the Coin reduction with defaults. Finally, it will train a model using the configuration file and the dataset. The training run will be evaluated using the mean squared error metric.

curl https://raw.githubusercontent.com/VowpalWabbit/vowpal_wabbit/master/test/test-sets/0001.dat > rcv1_small.vwtxt
reml config new Coin > config.json
reml train --config config.json --data rcv1_small.vwtxt --metrics mse

Output:

warning: This CLI tool is not stable
info: Reading data file: rcv1_small.vwtxt
info: Using entry reduction: Coin
+-----------------------------+
| Example #  MeanSquaredError |
+=============================+
| 0          0                |
| 1          0.5              |
| 2          0.33609757       |
| 4          0.20314142       |
| 8          0.11399492       |
| 16         0.19589658       |
| 32         0.14362814       |
| 64         0.15037861       |
| 100        0.13510321       |
+-----------------------------+