Package: tabnet 0.9.0
Christophe Regouby
tabnet: Fit 'TabNet' Models for Classification and Regression
Implements the 'TabNet' model by Sercan O. Arik et al. (2019) <doi:10.48550/arXiv.1908.07442> with 'Coherent Hierarchical Multi-label Classification Networks' by Giunchiglia et al. <doi:10.48550/arXiv.2010.10151> and provides a consistent interface for fitting and creating predictions. It's also fully compatible with the 'tidymodels' ecosystem.
Authors:
tabnet_0.9.0.tar.gz
tabnet_0.9.0.tar.gz(r-4.7-any)tabnet_0.9.0.tar.gz(r-4.6-any)
tabnet_0.9.0.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
card.svg |card.png
tabnet/json (API)
NEWS
| # Install 'tabnet' in R: |
| install.packages('tabnet', repos = c('https://cran.r-universe.dev', 'https://cloud.r-project.org')) |
Bug tracker:https://github.com/mlverse/tabnet/issues
Pkgdown/docs site:https://mlverse.github.io
Last updated from:6f51ecf80f. Checks:4 OK. Indexed: no.
| Target | Result | Time | Files | Syslog |
|---|---|---|---|---|
| linux-devel-x86_64 | OK | 233 | ||
| source / vignettes | OK | 242 | ||
| linux-release-x86_64 | OK | 229 | ||
| wasm-release | OK | 312 |
Exports:%>%attention_widthaugmentbuild_ancestor_matrix_from_outcomescat_emb_dimcheck_compliant_nodecheckpoint_epochsdecision_widthdrop_lastencoder_activationentmaxentmax15feature_reusagelr_schedulermask_typemlp_activationmlp_hidden_multipliermomentumnn_aum_lossnn_mc_lossnnf_mc_lossnnf_multilabel_one_hotnode_to_dfnum_independentnum_independent_decodernum_sharednum_shared_decodernum_stepsoptimizerpenaltysparsemaxsparsemax15tabnettabnet_configtabnet_explaintabnet_fittabnet_nntabnet_pretrainverbosevirtual_batch_size
Dependencies:base64encbitbit64bslibcachemcallrclasscliclockcodetoolscorocpp11crayondata.tabledata.treedescdiagramdialsDiceDesigndigestdplyrevaluatefarverfastmapfontawesomefsfurrrfuturefuture.applyGauProgenericsggplot2globalsgluegowergtablehardhathighrhmshtmltoolsipredisobandjquerylibjsonliteKernSmoothknitrlabelinglatticelavalbfgslifecyclelistenvlubridatemagrittrMASSMatrixmemoisemimemixoptmodelenvnnetnumDerivotelparallellyparsnippillarpkgconfigprettyunitsprocessxprodlimprogressprogressrpspurrrR6rappdirsRColorBrewerRcppRcppArmadillorecipesrlangrmarkdownrpartrsampleS7safetensorssassscalessfdshapeslidersparsevctrssplitfngrSQUAREMstringistringrsurvivaltailortibbletidyrtidyselecttimechangetimeDatetinytextorchtunetzdbutf8vctrsviridisLitewarpwithrworkflowsxfunyamlyardstickzeallot
Fitting tabnet with tidymodels
Rendered fromtidymodels-interface.Rmdusingknitr::rmarkdownon Jun 12 2026.Last update: 2026-01-31
Started: 2021-01-14
Hierarchical Classification
Rendered fromHierarchical_classification.Rmdusingknitr::rmarkdownon Jun 12 2026.Last update: 2026-06-12
Started: 2023-12-06
Interpretation tools
Rendered frominterpretation.Rmdusingknitr::rmarkdownon Jun 12 2026.Last update: 2026-01-31
Started: 2021-01-14
Self-supervised training and fine-tuning
Rendered fromselfsupervised_training.Rmdusingknitr::rmarkdownon Jun 12 2026.Last update: 2026-06-12
Started: 2023-12-06
Training a Tabnet model from missing-values dataset
Rendered fromMissing_data_predictors.Rmdusingknitr::rmarkdownon Jun 12 2026.Last update: 2025-04-17
Started: 2023-05-11
Using ROC AUM loss for imbalanced binary classification
Rendered fromaum_loss.Rmdusingknitr::rmarkdownon Jun 12 2026.Last update: 2026-01-31
Started: 2026-01-31
Readme and manuals
Help Manual
| Help page | Topics |
|---|---|
| Parameters for the tabnet model | attention_width decision_width feature_reusage mask_type momentum num_independent num_shared num_steps |
| Plot tabnet_explain mask importance heatmap | autoplot.tabnet_explain |
| Plot tabnet_fit model loss along epochs | autoplot.tabnet_fit autoplot.tabnet_pretrain |
| Build ancestor matrix aligned with observed outcome classes | build_ancestor_matrix_from_outcomes |
| Non-tunable parameters for the tabnet model | cat_emb_dim checkpoint_epochs drop_last encoder_activation lr_scheduler mlp_activation mlp_hidden_multiplier num_independent_decoder num_shared_decoder optimizer penalty verbose virtual_batch_size |
| Check that Node object names are compliant | check_compliant_node |
| Alpha-entmax | entmax entmax15 |
| Apply hierarchy constraints via max-pooling over descendants (MCM) | get_constr_output |
| Optimal threshold (tau) computation for 1.5-entmax | get_tau |
| AUM loss | nn_aum_loss |
| Max-Constraint Margin Loss (module) | nn_mc_loss |
| Prune top layer(s) of a tabnet network | nn_prune_head.tabnet_fit nn_prune_head.tabnet_pretrain |
| Max-Constraint Margin Loss (functional) | nnf_mc_loss |
| Convert class_id tensor to binary one-hot tensor | nnf_multilabel_one_hot |
| Turn a Node object into predictor and outcome. | node_to_df |
| Predict using 'tabnet' | augment.tabnet_fit predict.tabnet_fit |
| Sparsemax | sparsemax sparsemax15 |
| Parsnip compatible tabnet model | tabnet |
| Configuration for TabNet models | tabnet_config |
| Interpretation metrics from a TabNet model | tabnet_explain tabnet_explain.default tabnet_explain.model_fit tabnet_explain.tabnet_fit tabnet_explain.tabnet_pretrain |
| Tabnet model | tabnet_fit tabnet_fit.data.frame tabnet_fit.default tabnet_fit.formula tabnet_fit.Node tabnet_fit.recipe |
| TabNet Model Architecture | tabnet_nn |
| Tabnet model | tabnet_pretrain tabnet_pretrain.data.frame tabnet_pretrain.default tabnet_pretrain.formula tabnet_pretrain.Node tabnet_pretrain.recipe |
