v0.4.2

bug fix

v0.4.1

Faster k-NN search and simplex weighted average on both CPU and CUDA (1.3-5 times, depending on shape and device), plus a new moran_matrix(...).

Full Changelog: v0.3.1...v0.4.1

Changes

• Added moran_matrix(...) to PairwiseCCM, computing Moran's I of target vectors over source k-NN graphs.
• Improved the k-NN search: matmul-based squared distances in a reusable workspace, with the clamp and square root applied only to the selected neighbors instead of the whole distance block.
• Replaced the full-block exclusion mask with a sparse one that touches only the columns a query can actually exclude.
• Reused the neighbor workspace between calls by default: clean_after in score_matrix(...) and predict_matrix(...) now defaults to False, matching the documented behavior (1.5-3 times faster for repeated calls).
• Improved the simplex weighted average by using embedding_bag for wide target blocks, avoiding a bmm slowdown that affected target embeddings with 5 or more columns (up to ~5 times faster on CPU, ~1.3 times on CUDA).
• Added a prefiltered neighbor selection on CUDA for libraries above 4096 points (~1.7 times faster for long series).
• Improved streaming metric accumulation and CUDA simplex batching.
• Added benchmark scripts and generated CPU/GPU reports (scripts/README_CPU.md, scripts/README_GPU.md).

FastCCM v0.3.1

Full Changelog: v0.2.4...v0.3.1

Changes

• Added batching across targets.
• Added configurable memory_budget_gb and improved auto-batching for simplex and SMAP.
• Improved simplex and SMAP performance, including target batching and optional SMAP precomputation.
• Added optional global-baseline subtraction via subtract_global in PairwiseCCM.
• Improved blocked Functions utilities with suffix alignment, progress bars, and more consistent tp handling.

FastCCM v0.2.4

Full Changelog: v0.2.3...v0.2.4

Changes

• Improved simplex weighted-average performance (~2 times faster).
• Improved batch processing via metric streaming (reduces peak RAM/VRAM use).
• Added logger, progress bar, per-batch timing logs.
• Improved OOM handling.
• Added optional disk-backed outputs (out_path) for Functions.score_matrix_blocked and Functions.predict_matrix_blocked using numpy.memmap.

FastCCM v0.2.3

Full Changelog: v0.2.2...v0.2.3

Changes

• Match results/tests with pyEDM.
• Fix SMAP implementation to prevent memory/compute blow-up and reduce unnecessary computation.

FastCCM v0.2.2

Full Changelog: v0.2.0...v0.2.2

• Out-of-memory (OOM) handling.
• Chunked processing for memory safety.
• Cleanup on completion and on error.
• Consistent dtype handling.

FastCCM v0.2.1

Full Changelog: v0.2.0...v0.2.1

• Robust memory safety. Automatic batching for both simplex and smap paths plus cleanup.
• Switches to Cholesky solves instead of pseudo-inverse in SMAP.
• Cleaner public API & docs

FastCCM v0.2.0

Full Changelog: https://github.com/z7743/FastCCM/commits/v0.2.0

FastCCM v0.2.0 — Public API

fastccm

• class PairwiseCCM(...)
  • score_matrix(...)
  • predict_matrix(...)

fastccm.ccm_utils

• class Functions(...)
  • score_matrix_blocked(...)
  • predict_matrix_blocked(...)
  • convergence_test(...)
  • prediction_interval_test(...)
  • find_optimal_embedding_params(...)
• class Visualizer(...)
  • plot_convergence_test(...)
  • visualize_optimal_e_tau(...)
  • plot_interval_prediction_test(...)

fastccm.utils

• embed(...)
• get_td_embedding_np(...)
• calculate_correlation_dimension(...)
• calculate_rank_for_variance(...)
• metrics.get_metric(...)

fastccm.data

• get_truncated_lorenz_rand(...)
• get_truncated_rossler_lorenz_rand(...)