Algorithm Overview

The framework integrates algorithms from multiple recommendation system libraries through unified runner interfaces. Use the algorithm identifiers with the ExperimentPlan to configure experiments. Here is the comprehensive list of all available algorithms:

LensKit Algorithms

LensKit provides traditional collaborative filtering algorithms optimized for research reproducibility.

Algorithm Name	Type	Feedback	Description
LensKit.PopScorer	Baseline	Both	Popularity-based scorer
LensKit.ItemKNNScorer	Neighborhood	Both	Item-based k-nearest neighbors
LensKit.UserKNNScorer	Neighborhood	Both	User-based k-nearest neighbors
LensKit.ImplicitMFScorer	Matrix Factorization	Implicit	Implicit feedback matrix factorization (ALS)
LensKit.BiasedMFScorer	Matrix Factorization	Explicit	Biased matrix factorization
LensKit.FunkSVDScorer	Matrix Factorization	Explicit	FunkSVD matrix factorization

RecPack Algorithms

RecPack provides efficient implementations of collaborative filtering algorithms with a focus on scalability and performance.

Algorithm Name	Type	Feedback	Description
RecPack.SVD	Matrix Factorization	Explicit	Singular Value Decomposition
RecPack.NMF	Matrix Factorization	Explicit	Non-negative Matrix Factorization
RecPack.ItemKNN	Neighborhood	Both	Item-based k-nearest neighbors

Elliot Algorithms

Elliot provides a comprehensive framework for reproducible recommender systems evaluation with implementations of state-of-the-art algorithms.

Baseline and Neighborhood Methods

Algorithm Name	Type	Feedback	Description
Elliot.MostPop	Baseline	Both	Most popular items baseline
Elliot.ItemKNN	Neighborhood	Both	Item-based k-nearest neighbors
Elliot.UserKNN	Neighborhood	Both	User-based k-nearest neighbors
Elliot.SlopeOne	Slope One	Explicit	Slope One collaborative filtering

Matrix Factorization

Algorithm Name	Type	Feedback	Description
Elliot.AMF	Matrix Factorization	Explicit	Alternating Matrix Factorization
Elliot.BPRMF	Matrix Factorization	Implicit	Bayesian Personalized Ranking Matrix Factorization
Elliot.BPRMF_batch	Matrix Factorization	Implicit	BPRMF with batch processing
Elliot.FM	Factorization Machine	Explicit	Factorization Machine
Elliot.FunkSVD	Matrix Factorization	Explicit	FunkSVD matrix factorization
Elliot.NonNegMF	Matrix Factorization	Explicit	Non-negative Matrix Factorization
Elliot.PureSVD	Matrix Factorization	Explicit	Pure SVD
Elliot.SVDpp	Matrix Factorization	Explicit	SVD++
Elliot.WRMF	Matrix Factorization	Implicit	Weighted Regularized Matrix Factorization
Elliot.ConvMF	Matrix Factorization	Explicit	Convolutional Matrix Factorization
Elliot.DeepFM	Deep Learning	Explicit	Deep Factorization Machine
Elliot.DMF	Matrix Factorization	Explicit	Deep Matrix Factorization
Elliot.GMF	Matrix Factorization	Explicit	Generalized Matrix Factorization
Elliot.NeuMF	Deep Learning	Implicit	Neural Matrix Factorization

Autoencoder-Based Methods

Algorithm Name	Type	Feedback	Description
Elliot.MultiDAE	Deep Learning	Implicit	Multi-layer Denoising Autoencoder
Elliot.MultiVAE	Deep Learning	Implicit	Multi-layer Variational Autoencoder
Elliot.ItemAutoRec	Deep Learning	Explicit	Item-based Autoencoder
Elliot.UserAutoRec	Deep Learning	Explicit	User-based Autoencoder

Graph-Based Methods

Algorithm Name	Type	Feedback	Description
Elliot.LightGCN	Graph Neural Network	Implicit	Light Graph Convolutional Network
Elliot.NGCF	Graph Neural Network	Implicit	Neural Graph Collaborative Filtering

RecBole Algorithms

RecBole provides a comprehensive collection of modern recommendation algorithms including deep learning and graph-based methods.

Baseline and Neighborhood Methods

Algorithm Name	Type	Feedback	Description
RecBole.Pop	Baseline	Both	Popularity-based recommender
RecBole.ItemKNN	Neighborhood	Both	Item-based k-nearest neighbors
RecBole.Random	Baseline	Both	Random recommendation baseline

Matrix Factorization

Algorithm Name	Type	Feedback	Description
RecBole.BPR	Matrix Factorization	Implicit	Bayesian Personalized Ranking
RecBole.FISM	Matrix Factorization	Implicit	Factored Item Similarity Models
RecBole.NAIS	Matrix Factorization	Implicit	Neural Attentive Item Similarity
RecBole.DMF	Matrix Factorization	Explicit	Deep Matrix Factorization
RecBole.ENMF	Matrix Factorization	Implicit	Efficient Neural Matrix Factorization
RecBole.NNCF	Matrix Factorization	Implicit	Neural Network Collaborative Filtering

Deep Learning Methods

Algorithm Name	Type	Feedback	Description
RecBole.NeuMF	Deep Learning	Implicit	Neural Matrix Factorization
RecBole.ConvNCF	Deep Learning	Implicit	Convolutional Neural Collaborative Filtering
RecBole.CDAE	Deep Learning	Implicit	Collaborative Denoising Auto-Encoder
RecBole.MultiVAE	Deep Learning	Implicit	Variational Autoencoders for Collaborative Filtering
RecBole.MultiDAE	Deep Learning	Implicit	Denoising Autoencoders for Collaborative Filtering
RecBole.MacridVAE	Deep Learning	Implicit	Macroscopic and Microscopic Variational Autoencoder
RecBole.RecVAE	Deep Learning	Implicit	Variational Autoencoders for Recommendations
RecBole.DiffRec	Deep Learning	Implicit	Diffusion Recommender Model
RecBole.LDiffRec	Deep Learning	Implicit	Latent Diffusion Recommender Model

Graph-Based Methods

Algorithm Name	Type	Feedback	Description
RecBole.SpectralCF	Graph Neural Network	Implicit	Spectral Collaborative Filtering
RecBole.GCMC	Graph Neural Network	Explicit	Graph Convolutional Matrix Completion
RecBole.NGCF	Graph Neural Network	Implicit	Neural Graph Collaborative Filtering
RecBole.LightGCN	Graph Neural Network	Implicit	Light Graph Convolutional Network
RecBole.DGCF	Graph Neural Network	Implicit	Disentangled Graph Collaborative Filtering
RecBole.SGL	Graph Neural Network	Implicit	Self-supervised Graph Learning
RecBole.NCL	Graph Neural Network	Implicit	Neighborhood-enriched Contrastive Learning
RecBole.LINE	Graph Embedding	Implicit	Large-scale Information Network Embedding

Linear and Optimization-Based Methods

Algorithm Name	Type	Feedback	Description
RecBole.EASE	Linear	Implicit	Embarrassingly Shallow Autoencoders
RecBole.SLIMElastic	Linear	Implicit	Sparse Linear Method with ElasticNet
RecBole.ADMMSLIM	Linear	Implicit	ADMM SLIM for Top-N Recommendation
RecBole.NCEPLRec	Linear	Implicit	Neighborhood-based Collaborative Filtering with Pairwise Learning
RecBole.SimpleX	Linear	Implicit	Simple and Effective Collaborative Filtering

Using Algorithms in Experiments

Reference algorithms using the format <Runner>.<Algorithm> in your experiment plan:

from omnirec.runner.plan import ExperimentPlan
from omnirec.runner.algos import LensKit, RecBole

# Create experiment plan
plan = ExperimentPlan("Algorithm-Comparison")

# Add LensKit algorithm
plan.add_algorithm(
    LensKit.ItemKNNScorer,
    {"max_nbrs": 20, "min_nbrs": 5}
)

# Add RecBole algorithm
plan.add_algorithm(
    RecBole.LightGCN,
    {"embedding_size": 64, "n_layers": 3}
)

See the Algorithm Configuration guide for detailed information on configuring algorithms and hyperparameters.