Skip to main content

library to compute fermat distance

Project description

Fermat distance

Fermat is a Python library that computes the Fermat distance estimator (also called d-distance estimator) proposed in

Table of contents

  1. Introduction
  2. Instalation
  3. Implementation
  4. Features
  5. Support
  6. Citting Fermat distance


A density-based estimator for weighted geodesic distances is proposed. Let M be a D-dimensional manifold and consider a sample of N points X_n living in M. Let l(.,.) be a distance defined in M (a typical choice could be Euclidean distance). For d>=1 and given two points p and q in M we define the Fermat distance estimator as

The minimization is done over all K>=2 and all finite sequences of data points with x1= argmin l(x,p) and xK = argmin l(x,q).

When d=1, we recover the distance l(.,.) but if d>1, the Fermat distance tends to follow more closely the manifold structure and regions with high density values.


The optimization performed to compute the Fermat distance estimator runs all over the possible paths of points between each pair of points. We implement an algorithm that computes the exact Fermat distance and two that compute approximations.

  • Exact: Floyd-Warshall

Permorf the Floyd-Warshall algorithm that gives the exact Fermat distance estimator in O( n^3 ) operations between all possible paths that conects each pair of points.

  • Aprox: Dijsktra + k-nearest neighbours

With probability arbitrary high we can restrict the minimum path search to paths where each consecutive pair of points are k-nearest neighbours, with k = O(log n). Then, we use Dijkstra algorithm on the graph of k-nearest neighbours from each point. The complexity is O( n * ( k * n * log n ) ).

  • Aprox: Landmarks

If the number of points n is too high and neither Floyd-Warshall and Dijkstra run in appropiate times, we implemente a gready version based on landmarks. Let consider a set of l of point in the data set (the landmarks) and denote s_j the distance of the point s to the landmark j. Then, we can bound the distance d(s,t) between any two points s and t as

lower = max_j { | s_j - t_j | } <= d(s,t) <= min_j { s_j + t_j } = upper

and estimate d(s,t) as a function of lower and upper (for example, d(s,t) ~ (_lower + upper_) / 2 ). The complexity is O( l * ( k * n * log n ) ).



If you have an open-ended or a research question:

  • ''

Citting Fermat distance

When citing fermat in academic papers and theses, please use this BibTeX entry:

      title={Weighted Geodesic Distance Following Fermat's Principle},
      author={Facundo Sapienza and Pablo Groisman and Matthieu Jonckheere},

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for fermat, version 0.0.3
Filename, size File type Python version Upload date Hashes
Filename, size fermat-0.0.3-py3-none-any.whl (10.8 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size fermat-0.0.3.tar.gz (219.9 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page