Skip to content
geomstats-fisher-rao <<<<<<< HEAD logo

geomstats-fisher-rao <<<<<<< HEAD

Information geometry for Bayesian inference. Fisher-Rao metric on statistical manifolds, Riemannian optimization for model selection, Wasserstein geodesics connecting geomstats to Gromov-Wasserstein optimal transport.

plurigrid/asi0installs67stars

SKILL.md

Full skill instructions

geomstats-fisher-rao Skill

The natural metric on probability distributions is the Fisher-Rao metric

What It Covers

======= description: > Information geometry for Bayesian inference via geomstats. Triggers: Fisher-Rao metric, statistical manifolds, Riemannian optimization, Wasserstein geodesics, Gromov-Wasserstein optimal transport, hyperbolic embeddings, SPD matrices.

geomstats-fisher-rao

Information geometry connecting geomstats to Bayesian inference and optimal transport.

Coverage

origin/​main

22 notebooks in plurigrid's geomstats fork spanning:

  • Information geometry (Fisher information matrix as Riemannian metric)
  • Hyperbolic embeddings (Poincare ball, hyperboloid)
  • Graph space (Frechet mean of graphs)
  • Heisenberg group geometry
  • SPD matrices (covariance manifolds)

monad-bayes Connection

<<<<<<< HEAD The Fisher-Rao metric is the unique Riemannian metric (up to scale) that is invariant under sufficient statistics. This means:

MonadDistribution m => FisherRao (distribution m)

Every score/factor call in monad-bayes implicitly moves along a geodesic on the Fisher-Rao manifold. The MCMC acceptance ratio is the exponential map.

The Fisher-Rao metric is the unique Riemannian metric (up to scale) invariant under sufficient statistics:

origin/​main

-- Natural gradient MCMC on Fisher-Rao manifold
naturalGradientStep :: MonadMeasure m => SPDMatrix -> m Parameters
naturalGradientStep fisherInfo = do
  currentParams <- get
  proposal <- mvNormal currentParams (inverse fisherInfo)
  let logRatio = logLikelihood proposal - logLikelihood currentParams
  accept <- bernoulli (min 1 (exp logRatio))
  if accept then return proposal else return currentParams

<<<<<<< HEAD

Gromov-Wasserstein Bridge

geomstats <-> plurigrid/​ontology:GW via optimal transport:

  • Wasserstein distance = geodesic distance on the space of measures
  • Gromov-Wasserstein = comparison of metric measure spaces
  • Entropic regularization = softmax (connects to monad-bayes softmax)
  • Bregman projections for marginal constraints

Applications

  • Cortical manifold geometry (zubyul/​Nikolova_lab)
  • Protein structure manifolds (Vertex AI)
  • Eye movement geometry on the visual sphere (zubyul/​EyeGestures)
  • Color gamut as Riemannian manifold (Gay.jl DeltaE2000)

GF(3) Trit Classification

ComponentTritRole
geomstats computation+1Generation (geometric objects)
Fisher-Rao metric0Coordination (natural metric)
Wasserstein validation-1Validation (distance bounds)

Conservation: +1 + 0 + (-1) = 0

Trit: 0 (ERGODIC)

======= Every score/factor call in monad-bayes implicitly moves along a geodesic on the Fisher-Rao manifold. The MCMC acceptance ratio is the exponential map.

Gromov-Wasserstein Bridge

geomstats connects to Gromov-Wasserstein optimal transport:

  • Wasserstein distance = geodesic distance on the space of measures
  • Gromov-Wasserstein = comparison of metric measure spaces
  • Entropic regularization = softmax (connects to monad-bayes)
  • Bregman projections for marginal constraints

Applications

  • Cortical manifold geometry
  • Protein structure manifolds (Vertex AI)
  • Eye movement geometry on the visual sphere
  • Color gamut as Riemannian manifold (DeltaE2000)

Concrete Affordance: Runnable Fisher-Rao Distance Computation

Install

pip install geomstats
# geomstats pulls in numpy, scipy, matplotlib

Plurigrid Fork

# Clone the plurigrid fork (already present locally):
git clone [email protected]:plurigrid/​geomstats.git

# Upstream:
# git clone [email protected]:geomstats/​geomstats.git

Local fork path: /​Users/​alice/​v/​worlds/​g/​geomstats

Notebooks (21 in the fork)

#TopicPath
00Introduction to geomstatsnotebooks/​00_foundations__introduction_to_geomstats.ipynb
02Connection & Riemannian metricnotebooks/​02_foundations__connection_riemannian_metric.ipynb
08Information geometry (Fisher-Rao)notebooks/​08_practical_methods__information_geometry.ipynb
12EMG classification on SPD manifoldnotebooks/​12_real_world_applications__emg_sign_classification_in_spd_manifold.ipynb
13Graph embedding in hyperbolic spacenotebooks/​13_real_world_applications__graph_embedding_and_clustering_in_hyperbolic_space.ipynb
20Graph space (Frechet mean)notebooks/​20_real_world_applications__graph_space.ipynb
21Heisenberg group geometrynotebooks/​21_foundations__sub_riemannian_geometry_and_the_heisenberg_group.ipynb

All notebooks at: /​Users/​alice/​v/​worlds/​g/​geomstats/​notebooks/

Runnable: Fisher-Rao Distance Between Two Normal Distributions

#!/​usr/​bin/​env python3
"""
Compute Fisher-Rao geodesic distance between two univariate normal
distributions on the statistical manifold.

The Fisher information matrix for N(mu, sigma^2) is:
  g = diag(1/​sigma^2, 2/​sigma^2)

The Fisher-Rao distance between N(mu1,s1^2) and N(mu2,s2^2) is the
geodesic distance on the upper half-plane with this metric.

Run:
  python3 fisher_rao_demo.py
"""

import geomstats.backend as gs
gs.random.seed(42)

from geomstats.information_geometry.normal import NormalDistributions

# The statistical manifold of univariate normals
manifold = NormalDistributions()

# Two distributions: N(0, 1) and N(3, 2)
# Parameters are [mu, sigma] (NOT sigma^2)
p = gs.array([0.0, 1.0])   # N(0, 1)
q = gs.array([3.0, 2.0])   # N(3, 2)

# Fisher-Rao geodesic distance
dist = manifold.metric.dist(p, q)
print(f"Fisher-Rao distance between N(0,1) and N(3,2): {float(dist):.6f}")

# Geodesic curve (10 points from p to q on the statistical manifold)
geodesic_fn = manifold.metric.geodesic(initial_point=p, end_point=q)
t = gs.linspace(0.0, 1.0, 10)
geodesic_points = geodesic_fn(t)
print(f"\nGeodesic path (mu, sigma) from N(0,1) → N(3,2):")
for i, pt in enumerate(geodesic_points):
    print(f"  t={float(t[i]):.1f}: N({float(pt[0]):.3f}, {float(pt[1]):.3f})")

# Fisher information matrix at a point
fim = manifold.metric.metric_matrix(p)
print(f"\nFisher information matrix at N(0,1):\n{fim}")

SPD Manifold Quick Check

from geomstats.geometry.spd_matrices import SPDMatrices
import geomstats.backend as gs

spd = SPDMatrices(n=3)
A = spd.random_point()
B = spd.random_point()
d = spd.metric.dist(A, B)
print(f"Affine-invariant distance between two 3x3 SPD matrices: {float(d):.4f}")

Key Files

PathDescription
/​Users/​alice/​v/​worlds/​g/​geomstats/Plurigrid fork of geomstats
notebooks/​08_practical_methods__information_geometry.ipynbFisher-Rao notebook
geomstats/​information_geometry/​normal.pyNormalDistributions manifold implementation
geomstats/​geometry/​spd_matrices.pySPD matrices with affine-invariant metric

origin/​main

More skills from plurigrid

cargo logo
plurigrid/asi

cargo

Rust package manager (36 subcommands).

67 0
View

Translate Figma nodes into production-ready code with 1:1 visual fidelity using the Figma MCP workflow (design context, screenshots, assets, and project-convention translation). Trigger when the user provides Figma URLs or node IDs, or asks to implement designs or components that must match Figma...

67 0
View

**Status**: 🌳 Production Ready (upstream in StringZilla v3+) **Type**: High-Performance String Operations / SIMD-Accelerated Text Processing **Principle**: Zero-copy views, SIMD/SWAR acceleration, deterministic hashing **Frame**: Cross-language interoperability (C, C++, Python, Rust, Go, Swift, ...

67 0
View
conformal-ga logo
plurigrid/asi

conformal-ga

Conformal Geometric Algebra (CGA) for circles, spheres, and Möbius transformations

67 0
View

Verifies code implements exactly what documentation specifies for blockchain audits. Use when comparing code against whitepapers, finding gaps between specs and implementation, or performing compliance checks for protocol implementations.

67 0
View
zig-systems logo
plurigrid/asi

zig-systems

Systems programming and performance optimization using Zig. Provides low-level abstractions, memory-safe compiled code, and performance benchmarking. Use for system-level operations, optimization, and interoperability with C/POSIX APIs.

67 0
View
soliton-detection logo
plurigrid/asi

soliton-detection

Topological soliton detection and agency bridge with anyonic fusion algebra for concept composition

67 0
View
paypal-mcp logo
plurigrid/asi

paypal-mcp

PayPal MCP server integration for invoices, payments, subscriptions, disputes, and transaction reporting via @paypal/mcp.

67 0
View
unison logo
plurigrid/asi

unison

Unison language - content-addressed functional programming with abilities for effects, distributed computing, and structural types. Use for pure functional code, effect management, distributed systems, and refactoring-safe codebases.

67 0
View
tasks-acset logo
plurigrid/asi

tasks-acset

Google Tasks management via TasksACSet. Transforms task operations into GF(3)-typed Interactions, routes to triadic queues, detects saturation for task-zero-as-condensed-state.

67 0
View
c logo
plurigrid/asi

c

World C Skill

67 0
View
syrup logo
plurigrid/asi

syrup

Syrup binary serialization for OCapN/CapTP wire format. Canonical encoding for capability messages.

67 0
View

Popular AI tools

Kaiber logo
Video

Kaiber

Generate, edit, and beat-sync AI video with leading models in one workspace.

Paid
View
Vimcal logo
Productivity

Vimcal

The world's fastest calendar for remote work

Free
View

Transform Your Design with AI Designer by ImgCreator.ai

Freemium
View
Akool AI logo
Content & writing

Akool AI

Revolutionizing Video Production with AI-Powered Creativity

Paid
View

Extend an image past the frame and let AI fill the new aspect ratio.

Freemium
View
StarByFace logo
Security

StarByFace

Discover your celebrity doppelgänger with StarByFace!

Free
View
C

ChainClarity explains 700+ crypto whitepapers in plain English, with layered summaries, comparisons, research tools, alerts, and a $4.99 Pro plan.

Freemium
View
Opus Clip logo
Coding & apps

Opus Clip

Opus.ai: Revolutionize Your Web Experience

Free
View