LIBMF Rust

LIBMF - large-scale sparse matrix factorization - for Rust

Installation

Add this line to your application’s Cargo.toml under [dependencies]:

libmf = "0.1"

Getting Started

Prep your data in the format row_index, column_index, value

let mut data = libmf::Matrix::new();
data.push(0, 0, 5.0);
data.push(0, 2, 3.5);
data.push(1, 1, 4.0);

Fit a model

let model = libmf::Model::params().fit(&data).unwrap();

Make predictions

model.predict(row_index, column_index);

Get the latent factors (these approximate the training matrix)

model.p_factors();
model.q_factors();

Get the bias (average of all elements in the training matrix)

model.bias();

Save the model to a file

model.save("model.txt").unwrap();

Load the model from a file

let model = libmf::Model::load("model.txt").unwrap();

Pass a validation set

let model = libmf::Model::params().fit_eval(&train_set, &eval_set).unwrap();

Cross-Validation

Perform cross-validation

let avg_error = libmf::Model::params().cv(&data, 5).unwrap();

Parameters

Set parameters - default values below

libmf::Model::params()
    .loss(0)                // loss function
    .factors(8)             // number of latent factors
    .threads(12)            // number of threads used
    .bins(25)               // number of bins
    .iterations(20)         // number of iterations
    .lambda_p1(0.0)         // coefficient of L1-norm regularization on P
    .lambda_p2(0.1)         // coefficient of L2-norm regularization on P
    .lambda_q1(0.0)         // coefficient of L1-norm regularization on Q
    .lambda_q2(0.1)         // coefficient of L2-norm regularization on Q
    .learning_rate(0.1)     // learning rate
    .alpha(0.1)             // importance of negative entries
    .c(0.0001)              // desired value of negative entries
    .nmf(false)             // perform non-negative MF (NMF)
    .quiet(false);          // no outputs to stdout

Loss Functions

For real-valued matrix factorization

• 0 - squared error (L2-norm)
• 1 - absolute error (L1-norm)
• 2 - generalized KL-divergence

For binary matrix factorization

• 5 - logarithmic error
• 6 - squared hinge loss
• 7 - hinge loss

For one-class matrix factorization

• 10 - row-oriented pair-wise logarithmic loss
• 11 - column-oriented pair-wise logarithmic loss
• 12 - squared error (L2-norm)

Metrics

Calculate RMSE (for real-valued MF)

model.rmse(&data);

Calculate MAE (for real-valued MF)

model.mae(&data);

Calculate generalized KL-divergence (for non-negative real-valued MF)

model.gkl(&data);

Calculate logarithmic loss (for binary MF)

model.logloss(&data);

Calculate accuracy (for binary MF)

model.accuracy(&data);

Calculate MPR (for one-class MF)

model.mpr(&data, transpose);

Calculate AUC (for one-class MF)

model.auc(&data, transpose);

Reference

Specify the initial capacity for a matrix

let mut data = libmf::Matrix::with_capacity(3);

Resources

• LIBMF: A Library for Parallel Matrix Factorization in Shared-memory Systems

Upgrading

0.2.0

Use

let model = libmf::Model::params().factors(20).fit(&data).unwrap();

instead of

let mut model = libmf::Model::new();
model.factors = 20;
model.fit(&data);

History

View the changelog

Contributing

Everyone is encouraged to help improve this project. Here are a few ways you can help:

• Report bugs
• Fix bugs and submit pull requests
• Write, clarify, or fix documentation
• Suggest or add new features

To get started with development:

git clone --recursive https://github.com/ankane/libmf-rust.git
cd libmf-rust
cargo test