Skip to content

Changelog

All notable changes to RESOLVE will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.4.0] - 2025-01-25

Added

  • R² metric: Coefficient of determination for regression evaluation (computed on original scale)
  • Class weights: Support for imbalanced classification via class_weights in target config
  • LR scheduling: StepLR and CosineAnnealing scheduler options
  • Comprehensive test suite: Catch2 tests for C++, pytest for Python, testthat for R
  • CI/CD workflows: GitHub Actions for tests, PyPI release, R-CMD-check

Changed

  • Package split: C++ bindings now distributed as resolve-core, high-level Python wrapper as resolve
  • R package version synced to 0.4.0

Fixed

  • R² now computed on original scale (before inverse transform)
  • NAMESPACE exports for R metric functions

[0.2.0] - 2025-01-24

Changed

  • Simplified API: Trainer now takes only dataset as required argument, builds model automatically
  • Renamed normalization modes: relative_plot → relative, log_scaled → log
  • Loss presets: New loss_config parameter with presets "mae", "combined", "smape"
  • Confidence filtering: trainer.predict(dataset, confidence_threshold=0.8) to filter uncertain predictions

Added

  • Species encoding modes: species_encoding="hash" (default) or "embed" for learned species embeddings
  • Tensor caching: cache_dir parameter for faster restarts with large datasets
  • Gradient checkpointing: Reduced memory usage for large models

Fixed

  • Fixed expm1 overflow in metrics computation for extreme predictions
  • Improved R bindings with formula API support

[0.1.0] - 2025-01-19

Added

  • Initial release of RESOLVE (Representation Encoding for Structured Observation Learning with Vector Embeddings)
  • Hybrid species encoding: Feature hashing for full species lists + learned embeddings for dominant taxa
  • Multi-target prediction: Single shared encoder, multiple task heads (regression and classification)
  • Phased training: MAE -> SMAPE -> band accuracy optimization for regression targets
  • Semantic role mapping: Flexible column naming with strict structural requirements
  • Unknown species tracking: Detects and quantifies novel species at inference time
  • Abundance normalization: Raw, relative (per-plot), or log-scaled modes
  • CPU-first design: Works without GPU, scales with CUDA when available

Core Components

  • ResolveDataset: Data loading with semantic role mapping
  • ResolveModel: Neural network architecture with shared encoder and task-specific heads
  • Trainer: Training loop with phased optimization and early stopping
  • Predictor: Inference interface with embedding extraction