Release v0.7.0 · TomGeorge1234/SIMPL

Changes since v0.6.0

New features

Head direction cell (HDC) support: bundled headdirectioncells_vollan2025.npz dataset from Vollan et al. (2025), polar receptive field plotting, find_time_jumps() utility
Kalman filter now loops over all trials in a single pass (much faster with many trials)
load_demo_data() searches all GitHub releases (not just latest) for data files, and uses GITHUB_TOKEN in CI to avoid rate limits
Early stopping, bits-per-spike metric, slow speed prior warning
Built-in diagnostic plotting (plot_fitting_summary, plot_latent_trajectory, plot_receptive_fields, plot_all_metrics, plot_prediction)
Automatic docs website (mkdocs)
accumulate_spikes() utility, spatial information rate

API / breaking changes

sklearn-style refactor: config-only __init__, fit()/predict() API, trailing _ for fitted attributes
behaviour renamed to behavior throughout
epoch kwarg renamed to epochs
Removed verbose, prepare_dataset(), resample_spike_mask, observation_noise, delta_trajectory
Private methods made properly private

Housekeeping

Git history purged of large .npz data files, old notebook outputs, mkdocs build artifacts, and test PNGs (repo: 59 MB -> ~10 MB). If you cloned before 2025-03-24, re-clone the repo.
Data files now hosted on GitHub Releases and downloaded on first use
*.npz and data/ added to .gitignore
Auto-versioning via hatch-vcs from git tags

Bug fixes

FX memory bug, save/load round-trip bug, coarsen_dt bug, docs equation rendering, circular environment limits

fixed load_demo_data so it will look in all releases (not just latest… by @TomGeorge1234 in #67

Full Changelog: https://github.com/TomGeorge1234/SIMPL/commits/v0.7.0