Skip to content

03_cli_quickstart.md

Latest commit

 

History

History
60 lines (50 loc) · 1.94 KB

03_cli_quickstart.md

File metadata and controls

60 lines (50 loc) · 1.94 KB

Module 3: CLI Quickstart — Generate → Analyze → Visualize

  • Objectives
    • Use Mneme CLI to generate synthetic data and run the bioelectric pipeline
    • Explore topology backends and attractor options
  • Time: 45–60 minutes

3.1 Generate synthetic data

mneme generate -o data/synthetic/quickstart.npz -t bioelectric -s 64,64 --timesteps 10 --seed 7

3.2 Analyze (bioelectric defaults)

mneme analyze data/synthetic/quickstart.npz \
  --pipeline bioelectric \
  --topology-backend cubical \
  -o results_cli

Expected: results_cli/analysis_results.hdf5

3.3 Visualize dashboard

mneme visualize results_cli/analysis_results.hdf5 -o plots -f png

Expected: plots/dashboard.png with field, topology, and any attractor summaries.

If you prefer Python, you can also drive visualization programmatically:

from mneme.utils.io import load_results
from mneme.analysis.visualization import FieldVisualizer
from mneme.types import AnalysisResult, Field

ar = load_results('results_cli/analysis_results.hdf5')  # returns AnalysisResult
FieldVisualizer().create_analysis_dashboard(ar)

3.4 Backend and attractor variations

  • Rips (point-cloud):
mneme analyze data/synthetic/quickstart.npz \
  --pipeline bioelectric \
  --topology-backend rips \
  -o results_cli_rips
  • Disable attractors:
mneme analyze data/synthetic/quickstart.npz --attractor-method none -o results_noattr

Exercises

  1. Compare cubical vs rips outputs (feature vector length; diagram counts)
  2. Increase --attractor-threshold and note changes in detected attractors
  3. Try --attractor-method clustering with --attractor-min-samples 20

Solutions (outline)

  • Cubical operates directly on grids; Rips requires point-cloud conversion and may produce different diagram sparsity
  • Higher thresholds reduce recurrence connections → fewer attractors
  • Clustering groups dense regions; raising min_samples filters small basins