All the algorithms support sparse and dense matrix with multi card/node, unless otherwise specified.
| Category | Algorithm | Comment |
|---|---|---|
| Classification / Regression | Logistic Regression | |
| Multinomial Logistic Regression | ||
| Linear Regression | ||
| Ridge Regression | ||
| Lasso Regression | ||
| Linear Support Vector Classification | ||
| Linear Support Vector Regression | ||
| Kernel Support Vector Classification | [*1] | |
| Decision Tree Classification | [*3] | |
| Decision Tree Regression | [*3] | |
| Random Forest Classification | [*3] | |
| Random Forest Regression | [*3] | |
| GBDT Classification | [*3] | |
| GBDT Regression | [*3] | |
| Nearest Neighbor Classification | Only dense | |
| Nearest Neighbor Regression | Only dense | |
| Unsupervised Nearest Neighbor | Only dense | |
| Multinomial Naive Bayes | ||
| Bernoulli Naive Bayes | ||
| Clustering | K-means | |
| Spectral Clustering | Only dense | |
| Agglomerative Clustering | Only dense | |
| DBSCAN | Only dense | |
| ART-2A | Only C++ | |
| Preprocess | Singular Value Decomposition | |
| Eigen Value Decomposition | ||
| Principal Component Analysis | Only dense | |
| Spectral Embedding | Only dense | |
| T-SNE | Only dense | |
| Recommendation | ALS | |
| Factorization Machines | [*2] | |
| Basket Analysis | FP Growth | |
| Natural Language | word2vec | [*2] |
| Latent Dirichlet Allocation | ||
| Neural Network | Multi-layer Perceptron | Only C++ |
| Convolutional Neural Network | Only C++ | |
| Graph | PageRank | |
| Connected Components | ||
| Single-Source Shortest Path | ||
| Breadth First Search |
[*1] Only support OpenMP parallelization. Run with VE_OMP_NUM_THREADS=8 and mpirun -np 1
[*2] Support OpenMP + MPI parallelization. Better performance with VE_OMP_NUM_THREADS=8 and one MPI process per card.
[*3] Only dense matrix is supported, but category variables can be specified