AITCHISON GEOMETRY ON THE SIMPLEX FOR UNCERTAINTY QUANTIFICATION IN BAYESIAN HYPERSPECTRAL IMAGE UNMIXING
Hector Blondel, Lucas Drumetz, Thierry Chonavel · 2026
Read AITCHISON GEOMETRY ON THE SIMPLEX FOR UNCERTAINTY QUANTIFICATION IN BAYESIAN HYPERSPECTRAL IMAGE UNMIXING by Hector Blondel, Lucas Drumetz, Thierry Chonavel (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-24108.txt.
Introduction
Blondel and co-authors bring Aitchison geometry to uncertainty estimates in hyperspectral image unmixing. Unmixing finds the share of each material in every pixel, but most tools give only point estimates. Users then cannot tell how much to trust the shares. The authors use compositional geometry to build simplex-valued Gaussian process priors. They link this geometry to constrained sampling methods already in use. They also propose diagnostics that respect the sum constraints on abundance vectors. Tests use both real and simulated data.
Catalog record
| CREATOR | Hector Blondel, Lucas Drumetz, Thierry Chonavel |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27119 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.24108
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
AITCHISON GEOMETRY ON THE SIMPLEX FOR UNCERTAINTY QUANTIFICATION IN BAYESIAN HYPERSPECTRAL IMAGE UNMIXING Hector Blondel, Lucas Drumetz, Thierry Chonavel Abstract Most algorithms for hyperspectral image unmixing produce point estimates of fractional abundances of the materials to be separated. However, in the absence of reliable ground truth, the ability to perform abundance uncertainty quantification (UQ) should be an important feature of algorithms, e.g. to evaluate how hard the unmixing problem is and how much the results should be trusted. The usual modeling assumptions in Bayesian unmixing rely heavily on the Euclidean geometry of the simplex and typically disregard spatial information. In addition, to our knowledge, abundance UQ is close to nonexistent in the literature. In this paper, we propose to leverage Aitchison geometry used in compositional data analysis to provide practitioners with alternative tools for modeling prior abundance distributions. In particular, we show how to design simplex-valued Gaussian Process priors using this geometry. Then we link Aitchison geometry to constrained optimization and sampling algorithms, and propose UQ diagnostics that comply with the constraints on abundance vectors. We illustrate these concepts on real and simulated data. † † address: 1 IMT Atlantique, Lab-STICC, UMR CNRS 6285, Plouzané, France 2 INRIA ODYSSEY team-project, Brest, France 1 Introduction Hyperspectral image unmixing is a classical source separation problem in remote sensing [ 1 ] , wherein the spectral signatures of the materials present in the image (endmembers) and their relative proportions in each pixel (abundances) have to be estimated, with numerous applications in e.g. environmental monitoring [ 2 ] , urban planning [ 3 ] , planetology and astrophysics [ 4 ] . In spite of a plethora of algorithms taking into account difficulties such as the nonlinearity of the mixing phenomena [ 5 ] or the intra class variability of the endmembers [ 6 ] , the domain is plagued by the difficulty of rigorously validating algorithms on real data, due to the lack of reliable ground truth [ 7 ] . Simply assessing the difficulty of a specific instance of the problem performing uncertainty quantification (UQ) in the estimated quantities depends on many factors, ranging from the spatial and spectral resolutions, the granularity of the “pure” materials to separate, or the observation conditions. Even though a recent trend is to use data-driven probabilistic models to capture the variability of the endmembers (see e.g. [ 8 , 9 ] ), to our knowledge, the associated uncertainty on the abundances has not been explored or quantified extensively. The reasons for this may be that 1) Abundances are subject to constraints, making it harder to interpret uncertainty. Indeed, typically abundances belong to the unit simplex (though this can be relaxed in several ways) 2) The abundances are a 2D field, i.e. it is desirable and sometimes necessary (for noisy or incomplete data) to account for spatial information, which is not straightforward 3) Handling the posterior distribution is typically intractable and requires sophisticated sampling algorithms, in particular in the presence of constraints [ 10 , 11 ] . Thus, in this paper, we focus on modeling and representing the spatial and material-wise uncertainty of the abundances p(\mathbf{A}) , where \mathbf{A}\in\mathbb{R}^{P\times N} with P the number of materials and N the number of pixels. Each column \mathbf{a}_{n} of \mathbf{A} belongs to the probability (P-1) -simplex \Delta^{P-1}=\{\mathbf{a}\in[0,1]^{P}|\sum_{k=1}^{P}a_{k}=1\} . We work in an idealized Bayesian linear mixing model framework where we assume an endmember is represented by a single known signature: i.e. the likelihood of the observations \mathbf{X}\in\mathbb{R}^{L\times N} with L the number of spectral bands is given by p(\mathbf{X}|\mathbf{A})=\mathcal{N}(\mathbf{SA},\sigma^{2}\mathbf{I}_{L}) with \mathbf{S}\in\mathbb{R}^{L\times P} is the endmember matrix (containing the endmember signatures in its columns). With this setup, we aim to sample and represent the uncertainty of the posterior distribution: p(\mathbf{A}|\mathbf{X)}\propto p(\mathbf{X}|\mathbf{A})p(\mathbf{A}) (1) Dirichlet distributions are classically used as a natural model for the prior distribution of abundance vectors [ 12 ] . In compositional data analysis [ 13 ] , however, another geometry, so-called Aitchison geometry, is used because of favorable properties and natural handling of the constraints. To our knowledege, this geometry has not been considered in hyperspectral image unmixing: hence this paper investigates the potential benefits of such a geometry with a focus on a Bayesian view of the problem. We first review the basics of Aitchison geometry in a unmixing context. Then we show how it 1) allows to leverage bijections between \mathbb{R}^{P-1} and the interior of the simplex to define a new geometry that 2) can be used to extend Gaussian Processes (GP) to simplex-valued random fields by pushing forward Euclidean GP so that they respect the constraints. The resulting GP make for spatialized geometric priors allowing closed-form interpolation of abundance maps and provide sensible priors for spatialized Bayesian unmixing. 3) may be leveraged in optimization or sampling algorithms. 4) provide principled tools to define constraint-aware ways to represent and visualize abundance uncertainty at the pixel and image levels. 2 Aitchison Geometry In this section, we recall how diffeomorphic transformations \psi:\textrm{int}\ \Delta^{P-1}\to\mathbb{R}^{P-1} from compositional data analysis turn the interior of the simplex into a flat Riemannian manifold with a pullback Euclidean metric. This allows to design prior distributions on the simplex that are pushforwards of Gaussian distributions on \mathbb{R}^{P-1} . A straightforward choice for such a \psi is the alr transformation [ 13 ] \textrm{alr}(\mathbf{a}):=[\log a_{k}-\log a_{P}]_{1\leq k\leq P-1} , which removes one coordinate of \mathbf{a} to get rid of the sum-to-one constraint, and then takes the logarithm to alleviate the positivity constraint. However, the choice of the coordinate to remove is arbitrary, introducing symmetry-breaking artifacts in the representation. A way to symmetrize the previous transformation is the Centered Log Ratio (clr) [ 13 ] : \mathbf{w}=\psi(\mathbf{a})\triangleq\textrm{clr}(\mathbf{a}):=\begin{bmatrix}\log a_{p}-\frac{1}{P}\sum_{k=1}^{P}\log a_{k}\\ \end{bmatrix}_{p=1,...,P} (2) \textrm{clr}:\textrm{int}\ \Delta^{P-1}\to\mathbb{R}^{P-1} is not a bijection, but the restriction onto its image, the hyperplane H with normal vector \mathbf{1}\in\mathbb{R}^{P} , is a diffeomorphism. For numerical computations, choosing an orthonormal basis \mathbf{H}\in\mathbb{R}^{{P}\times{P-1}} of H allows to work in \mathbb{R}^{P-1} , and the resulting transformation is called the Isometric Log Ratio (ilr) \mathbf{z}=\textrm{ilr}(\mathbf{a})\triangleq\mathbf{H}^{T}\mathbf{w} . The inverse transformation is the softmax function: \textrm{ilr}^{-1}(\mathbf{z})=\textrm{softmax}(\mathbf{H}\mathbf{z})=\Bigg[\frac{\exp{w}_{p}}{\sum_{k=1}^{P}\exp(w_{k})}\Bigg]_{p=1,...,P}. (3) With a slight abuse of notation, we still refer to \textrm{ilr}(\mathbf{a}) as \psi(\mathbf{a}) . We note that \psi (as any diffeomorphism \textrm{int}\ \Delta^{P-1}\to\mathbb{R}^{P-1} ) turns \textrm{int}\ \Delta^{P-1} into a flat Riemannian manifold with at metric that is the pullback \psi^{*}g of the Euclidean metric g on \mathbb{R}^{P-1} . In other words, (\textrm{int}\ \Delta^{P-1},\psi^{*}g) is isometric to (\mathbb{R}^{P-1},g) by construction. The associated geodesic distance is given by the Euclidean distance in ilr space: d_{\psi}(\mathbf{a},\mathbf{a}^{\prime})=||\textrm{ilr}(\mathbf{a})-\textrm{ilr}(\mathbf{a}^{\prime})||_{2} (4) In particular, geodesics are obtained in the simplex by applying \textrm{ilr}^{-1} to straight lines in \mathbb{R}^{P-1} . This is a crucial difference from the usual Euclidean metric on \Delta^{P-1} that is induced from the Euclidean metric on \mathbb{R}^{P} . For the Euclidean metric, the geodesics are straight lines in the simplex. To illustrate the fundamental differences between both metrics, see Fig. 1 , where we plot both geodesics to joint two points of the simplex (P and Q) that correspond to the same abundances, up to a permutation of two of their components, and the other fixed. The Euclidean geodesic (in purple) is a straight line, and all abundances on the way will retain the same fraction of material (1,0,0)