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.

immunaut 1.0.3

• Machine Learning Data Leakage Fix: Enforced strict train/test data isolation in auto_simon_ml(). The dataset is partitioned into training and testing sets prior to any preprocessing transformations.
• Preprocessing Isolation: preProcessData() and preProcessResample() estimate imputation, centering, scaling, and zero-variance filtering parameters exclusively on the training partition and apply them to the test set without information leakage.
• Fitted Preprocessing Object: Added preProcessParams to the auto_simon_ml() output object for reproducible transformations on external validation datasets.
• Multi-Class Evaluation: Added MLmetrics to Imports and fixed multi-class Macro-F1 score calculation to correctly extract class-level metrics from confusionMatrix$byClass.
• Vectorized Metric Normalization in pick_best_cluster_overall(): Resolved a scalar normalization bug where candidate cluster scores were collapsed to 0.5. Candidate metrics (modularity, silhouette, Davies-Bouldin index, Calinski-Harabasz index) are now normalized across all candidate partitions simultaneously, with index alignment preserved.
• Robust NA Handling in normalize(): Handled NAs gracefully via range(..., na.rm = TRUE) to prevent missing value where TRUE/FALSE needed errors on incomplete vectors.
• Safe Comparison in pick_best_cluster_modularity() & pick_best_cluster_silhouette(): Added !is.na() checks to ensure clustering selection does not crash when candidates contain missing metric values.
• Clean Default Handling in immunaut(): Automatically generates fileHeader if NULL and defaults selectedColumns safely to avoid dropping all features.
• Preserved Rows in immunaut(..., removeNA = TRUE): Properly aligned dataset_ml rows with filtered observations so that clustering labels are retained even after na.omit() drops incomplete cases.
• Adaptive t-SNE and Louvain Parameter Clamping:
  • calculate_tsne() respects user-specified seeds (falling back to 1337) and dynamically clamps perplexity to the maximum mathematical threshold allowed by sample size ((n - 1) / 3).
  • Safe zero-variance feature filtering using isTRUE(stats::var(., na.rm = TRUE) > 0) prevents tidyselect predicate evaluation crashes on all-NA or constant columns.
  • cluster_tsne_knn_louvain() dynamically adjusts knn_clusters whenever \(k \ge n\) to prevent FNN::get.knn crashes and graph construction failures.
• DBSCAN and Outlier Factor Alignment: DBSCAN noise points (labeled 0 or NA) in cluster_tsne_density(), cluster_tsne_mclust(), and cluster_tsne_hierarchical() are properly mapped to cluster “100” with factor levels explicitly updated, eliminating accidental NA conversions.
• Plotting Improvements in plot_clustered_tsne():
  • Eliminated global R session modification caused by theme_set(), applying themes locally to the ggplot object.
  • Preserves non-numeric/string cluster identifiers as valid factors instead of coercing them to NA via as.numeric().
  • Removed stray debug print() statements in is_var_empty().
• Comprehensive Unit Testing: Added a test suite (test-auto_simon_ml.R and test-pipeline_and_fixes.R) with 43 assertions covering ML isolation, clustering methods, edge cases, NA robustness, and visualization.

immunaut 1.0.2

• Add datasets: Added new datasets - https://zenodo.org/records/14719593
• Add examples: Added examples for new datasets

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.