Experimental Analysis Examples

When you need to process real neural recordings or produce richer visualisations, start with the examples/experimental_* scripts. They show how to combine experimental_data, PlotConfig, and third-party tooling (UMAP, TDA, Numba, …) to build a full analysis pipeline.

experimental_cann1d_analysis.py

  • Location: examples/experimental_cann1d_analysis.py

  • Data source: load_roi_data() fetches ROI traces from the Hugging Face cache.

  • Analysis steps:

    1. Run bump_fits (MCMC) to extract bump parameters frame by frame.

    2. Configure CANN1DPlotConfig.for_bump_animation for title, frame rate, and brightness limits.

    3. Call create_1d_bump_animation to generate the GIF; nframes and progress-bar settings are optional tweaks.

  • Output: bump_analysis_demo.gif and summary statistics in the console.

  • Extensions:

    • Tune n_steps / n_roi to match your dataset.

    • Without Numba the script falls back to NumPy—watch the startup message if you care about runtime.

experimental_cann2d_analysis.py

  • Location: examples/experimental_cann2d_analysis.py

  • Data source: load_grid_data() downloads spike and position data for grid cells.

  • Analysis steps:

    1. Configure SpikeEmbeddingConfig (smoothing, speed filters) and run embed_spike_trains.

    2. Reduce dimensionality with umap.UMAP and visualise using plot_projection.

    3. Compute persistence with tda_vis + TDAConfig to validate torus topology.

    4. Decode phases via decode_circular_coordinates and animate with plot_3d_bump_on_torus.

  • Output: experimental_cann2d_analysis_torus.gif and supporting plots.

  • Extensions:

    • Set do_shuffle / num_shuffles in tda_config for statistical tests.

    • Use save_path arguments to archive projections and barcodes.

Tools & dependencies

  • On first run the helpers create ~/.canns/data and cache downloads automatically.

  • Additional libraries you may need: umap-learn, canns_lib.ripser (via canns-lib), numba, matplotlib. Install them with:

    uv add umap-learn canns-lib numba

  • PlotConfig and CANN2DPlotConfig accept show=False so you can render figures on headless servers without changing the code.