Getting started

Lenticulum learns relations instead of functions: one model of a joint space, queried in whichever direction you need. This page installs it, runs a first query, and points you to the tutorial that fits your background.

Install

In the MathStruct registry, next to General:

using Pkg
Pkg.Registry.add(RegistrySpec(url = "https://github.com/MathStruct/Registry.git"))   # once
Pkg.add("Lenticulum")             # the factors and inference
Pkg.add("VariationalDiffusion")   # relation learners

A registered version is a snapshot and falls behind the development branch. For the newest code, use a clone:

Pkg.develop(path = "/path/to/Lenticulum.jl/lib/Lenticulum.jl")
Pkg.develop(path = "/path/to/Lenticulum.jl/lib/VariationalDiffusion.jl")   # and the others under lib/

A first query

A diffusion model of points on the unit circle, in closed form so that nothing needs training, asked for $y$ given $x = 0.6$. The relation has two answers, and implicit_roots returns both:

using VariationalDiffusion, LuxCore, Random
sched = VPSDE()
θ = range(0, 2π; length = 49)[1:48]
circle = NoisePredictor(GaussianMixtureEps(sched, vcat(cos.(θ)', sin.(θ)'); s = 0.05), sched)
ps, st = LuxCore.setup(Xoshiro(0), circle)
m = ImplicitDiffusion(circle, field_nodes(Xoshiro(1), 2; samples = 8))

answers, _ = implicit_roots(m, [0.6, 0.0], [Inf, 0.0], ps, st)   # x clamped (Inf), y free (0)
[round(a.z[2]; digits = 3) for a in answers]
2-element Vector{Float64}:
  0.772
 -0.78

Swap the precision vector to [0.0, Inf] and the same model answers "$x$ given $y$".

Where to go next

you come fromstart with
machine learningA relation without training, then Train a small diffusion model, Robot arm: one model, every direction, the kernel baseline and Intersections
statistics or robotics (Kalman filters, GTSAM)Localisation as a factor graph: odometry and GPS, smoothing versus filtering, and the noise level estimated from the marginal likelihood
physicsSymmetry is not a law and Discovering a force law
category theorythe theory vault, entry point Factors are Parameterized Statistical Games
looking for an APIthe package pages, starting with LenticulumCore, and the Vocabulary

Every tutorial is also available as a Jupyter notebook, and every work cited is on the References page.