The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

srlars 1.0.0

• Initial stable release of package.

srlars 1.0.1

• Fix error related to exhaustion of number of predictors.

srlars 2.0.0

• Implementation of new algorithm for FSCRE.

srlars 2.0.1

• Fix README to reflect new version of srlars function.

srlars 3.0.0

• Breaking change: redesigned srlars() interface. Replaced robust=TRUE/FALSE with explicit preprocessing and CV controls (x_preprocess, y_preprocess, cor_estimator, cv_preprocess, cv_fit, cv_loss, cv_folds).
• No target leakage by default: predictor preprocessing is now performed on X only (no joint preprocessing of [y, X]).
• Added wrapping-based robust, PSD correlations (cor_estimator = "wrap"), based on cellWise::wrap().
• Added leakage-free foldwise CV preprocessing option (cv_preprocess = "foldwise") using cellWise::DDCpredict() and foldwise response transforms.
• Added robust arbiter options: Huber/trimmed/MSE scoring and optional Huber IRLS fitting inside the CV loop (cv_fit = "huber").
• Updated internal selection loop stopping logic to require strictly positive CV improvement before accepting a variable.

srlars 3.0.1

• Make cv_fit = "huber" the default.

srlars 3.1.0

• Fixed predict.srlars()’s dynamic DDC-cleaning of new data: object$robust is now set by srlars() (it was previously always missing, so the DDC-cleaning branch was dead code and dynamic = TRUE had no effect), and the stale dummy-response-column augmentation before calling cellWise::DDCpredict() was removed to match the fact that x_preprocess = "ddc" is fit on the predictors alone.
• Added max_share argument to srlars(): the maximum number of sub-models (1 to n_models) a given variable may appear in. Default is 1, reproducing the original fully-disjoint behavior exactly. When 1 < max_share < n_models, each sub-model’s first selected variable is forced to be distinct across sub-models, preventing several sub-models from redundantly duplicating the same strongest cold-start predictor; sharing is only permitted for variables added after a sub-model’s first pick. This restriction is lifted entirely at max_share = n_models (sub-models are then free to become identical).
• Added n_min argument to srlars(): the minimum number of variables each sub-model is guaranteed to receive (subject to availability), even if a candidate doesn’t clear the usual positive-benefit/tolerance requirement. Default is NULL (no floor, original behavior). Never bypasses the max_share/diversity pool restrictions – only the CV-benefit acceptance requirement is relaxed for sub-models below the floor.
• Added cv.srlars(): chooses max_share by (outer) cross-validation and returns the ensemble refit at the cross-validated optimum. coef() and predict() work directly on the result via the existing coef.srlars()/predict.srlars() methods.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.