release: v0.5.0

CHANGELOG.md

@@ -0,0 +1,71 @@
+# Changelog
+
+All notable changes to `RiskLabAI.jl` are documented here. The format is based on
+[Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and the project follows
+[Semantic Versioning](https://semver.org/spec/v2.0.0.html) (pre-1.0: minor
+versions may include breaking changes).
+
+## [0.5.0] — 2026-06-19
+
+First substantive release. The package was reconstructed from the ground up to
+mirror the `RiskLabAI.py` public API in idiomatic Julia (`snake_case` functions,
+`CapWords` types, sub-modules mirroring the Python sub-packages). Every
+deterministic function is verified against a Python reference; stochastic and
+machine-learning pieces are validated structurally. The parity ledger lives in
+`PARITY.md`.
+
+### Added
+
+- **Data**
+  - `Data.Structures` — standard, time, imbalance (expected/fixed) and run
+    (expected/fixed) bars.
+  - `Data.Differentiation` — fractional differentiation (standard + fixed-width).
+  - `Data.Weights` — sample uniqueness and return/time-decay weighting.
+  - `Data.Denoise` — Marčenko–Pastur denoising, `cov_to_corr`/`corr_to_cov`,
+    optimal-portfolio helpers.
+  - `Data.Labeling` — CUSUM filter, triple-barrier, meta-labeling, trend scanning.
+  - `Data.Distance` — variation of information, mutual information, KL/cross-entropy.
+  - `Data.SyntheticData` — `form_block_matrix`, `drift_volatility_burst`,
+    `compute_log_returns`, `align_params_length` (exact) and the Heston–Merton
+    regime-switching price simulators (behavioural).
+- **Features** — Shannon/plug-in/Lempel–Ziv/Kontoyiannis entropy; Corwin–Schultz
+  and Bekker–Parkinson microstructural estimators; ADF / (B)SADF structural
+  breaks; feature importance: `orthogonal_features` + `calculate_weighted_tau`
+  (exact) and `feature_importance_mdi`/`mda`/`sfi` + `get_test_dataset`
+  (DecisionTree.jl backend).
+- **Backtest** — backtest statistics, probabilistic Sharpe ratio, test-set
+  overfitting, strategy risk, probability of backtest overfitting, synthetic
+  backtesting, and bet sizing.
+- **Optimization** — Hierarchical Risk Parity building blocks + `hrp()`, PCA
+  hedging, and Nested Clustered Optimisation (`get_optimal_portfolio_weights`,
+  `get_optimal_portfolio_weights_nco`).
+- **Cluster** — Optimized Nested Clustering (`cluster_k_means_base`/`top`), an
+  exact `silhouette_samples`, and random block-correlation generators.
+- **Validation** — purged K-Fold (with embargo), Combinatorial Purged CV,
+  Walk-Forward and standard K-Fold splitters; `cross_val_score`; grid/random
+  hyper-parameter search (`grid_search_cv`/`random_search_cv`).
+- **Ensemble** — theoretical bagging accuracy (exact) plus an empirical
+  weighted-bagging evaluator and bootstrap accuracy.
+- **Pde** — the four Deep-BSDE equations (HJB-LQ, Black–Scholes–Barenblatt,
+  default-risk and different-rate pricing) with exact generators, and a neural
+  Deep-BSDE solver (`solve_deep_bsde`) on the Lux.jl backend.
+
+### Dependencies
+
+- Added `Clustering`, `DecisionTree`, `Lux`, `Zygote` and `Optimisers`.
+
+### Notes
+
+- The Python `controller`, `core`/registry and `hpc` infrastructure layers are
+  intentionally **not** ported; Julia's multiple dispatch and native
+  `Distributed` replace them (see `PARITY.md` for the full list of deliberate
+  divergences). Sample weights are unavailable in the DecisionTree.jl-backed
+  importances, and the Deep-BSDE set-transformer/Monte-Carlo/FBSNN variants are
+  research scaffolding that was not ported.
+
+## [0.0.1]
+
+- Initial package skeleton.
+
+[0.5.0]: https://github.com/RiskLabAI/RiskLabAI.jl/releases/tag/v0.5.0
+[0.0.1]: https://github.com/RiskLabAI/RiskLabAI.jl/releases/tag/v0.0.1

Project.toml

@@ -1,6 +1,6 @@
 name = "RiskLabAI"
 uuid = "a72881da-fdaa-49c1-8962-99caf4ccfee8"
-version = "0.0.1"
+version = "0.5.0"
 authors = ["RiskLab AI <research@risklab.ai>"]
 
 [deps]