myFM - Bayesian Factorization Machines in Python/C++¶
myFM is an unofficial implementation of Bayesian Factorization Machines in Python/C++. Notable features include:
Implementation of all corresponding functionalities in libFM MCMC engine (including grouping & relation block)
A simpler and faster implementation using Pybind11 and Eigen
Gibbs sampling for ordinal regression with probit link function. See the tutorial for its usage.
Support variational inference, which converges faster and requires lower memory (but usually less accurate than the Gibbs sampling).
In most cases, you can install the library from PyPI:
pip install myfm
It has an interface similar to sklearn, and you can use them for wide variety of prediction tasks. For example,
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn import metrics
from myfm import MyFMClassifier
dataset = load_breast_cancer()
X = StandardScaler().fit_transform(dataset['data'])
y = dataset['target']
X_train, X_test, y_train, y_test = train_test_split(
X, y, random_state=42
)
fm = MyFMClassifier(rank=2).fit(X_train, y_train)
print(metrics.roc_auc_score(y_test, fm.predict_proba(X_test)))
# 0.9954
Try out the following examples to see how Bayesian approaches to explicit collaborative filtering are still very competitive (almost unbeaten)!
Basic Usage
Details