Invariant Image Reparameterisation: Bridging Symbolic and Numerical Methods for Identifiability Analysis, Model Reduction, and Prediction

arXiv:2502.04867 · stat.AP · Submitted 2025-02-07 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.

Tom: I'm Tom.

Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.

Tom: We're looking at the paper: "Invariant Image Reparameterisation".

Jane: Here's the gist of the paper. Invariant Image Reparameterisation (IIR) is an approach to parameter identifiability analysis that links symbolic reparameterisation ideas with local numerical calculations to determine a basis-independent representation of parameter combinations associated with observable model behaviour.

Tom: This method is significant because it allows the symbolic reparameterisation condition to be replaced by a single numerical Jacobian calculation at a single reference point, bridging symbolic and numerical methods for identifiability analysis, model reduction, and prediction.

Jane: Core Concept of Invariant Image Reparameterisation The central object in this approach is the invariant image, which is defined as a reduced, basis-independent representation of the parameter combinations controlling observable model behaviour.

Tom: The authors show that when a one-to-one componentwise transformation makes observable behaviour depend only on fixed linear combinations of the transformed parameters, a single numerical Jacobian determines the associated lower-dimensional reparameterisation space.

Paper summary: Tom: To unpack that a bit more, the paper claims that when you have this one-to-one componentwise transformation, and the observable behavior only depends on fixed linear combinations of those transformed parameters, then a single numerical Jacobian calculation at one reference point is enough to determine the associated lower-dimensional reparameterisation space.

Jane: That sounds like it simplifies the symbolic condition into something we can actually compute with numbers, which is huge because symbolic methods often get stuck in idealized settings.

Lu: They show that this process finds a reduced parameter space called the invariant image, and they even mention that this image has multiple ways to be described, like using Singular Value Decomposition or sparse monomial bases.

Meng: That's interesting because if you can choose between the SVD basis and these interpretable monomial bases, it suggests we might get representations that are both mathematically sound and actually make sense for us when we look at the data.

Lalam: And they introduce a first-order invariance condition, which they use to tell the difference between minimal reductions, like finding the invariant part of the local null space in non-minimal cases.

Tom: So it's not just finding *any* reduction, but using that first-order condition to distinguish between different types of reductions—minimal versus non-minimal—which is a key distinction in identifiability analysis.

Jane: It’s like they are giving us a tool to decide if we're looking at the most fundamental set of parameters or just some partial simplification.

Lu: The symbolic condition itself is represented by differential conditions on the row and null spaces of an auxiliary mapping Jacobian, and they show how that can be exactly represented by local derivative information at a single reference point under certain structural assumptions.

Meng: That assumption about the transformed auxiliary mapping being written in terms of reduced transformed coordinates, like phi star theta star equals phi bar A theta star equals phi bar eta, is what lets them get this clean numerical form.

Lalam: The specific conditions they derive are that the row of A must contain the row of D theta star phi star theta stars, and the null space of A must be contained in the null space of D theta star phi star theta stars.

Conclusion: Tom: So what does this paper actually mean when we look at it from a broader perspective? The title Invariant Image Reparameterisation really captures that they are bridging the gap between symbolic reparameterisation ideas and actual numerical calculations for identifiability analysis, model reduction, and prediction.

Jane: It suggests that we don't always need those heavy, global symbolic calculations to know what parameters matter; sometimes a local numerical calculation at one spot gives you the same information up to some numerical tolerances.

Lu: The implication is that we can use these single-point Jacobian calculations to find reduced parameter spaces, even when we are dealing with models that might be structurally identifiable but practically weakly informed.

Meng: For deployment purposes, this could mean faster model reduction because instead of a global symbolic search, you just run one numerical check at the reference point to see what the relevant combinations are.

Lalam: It also opens up ways to choose which coordinates you use for that reduced space—like choosing between the default SVD construction and these more interpretable sparse monomial bases—which is a practical win for how we interpret our results.

Tom: So, essentially, they provide a numerical way to get information about identifiable parameter combinations without needing the full machinery of symbolic methods across the entire model structure.

Jane: It moves us closer to having practical tools for model reduction and prediction where we are dealing with real data and not just perfectly idealized mathematical setups.

Oliver J. Maclaren, Ruanui Nicholson, Joel A. Trent, Joshua Rottenberry, Matthew J. Simpson

University of Auckland · Queensland University of Technology

stat.AP

Submitted: 2025-02-07

Updated: 2026-11-08

Comments: 43 pages incl. supplementary material (main text approx. 30 pages). Accepted manuscript version of the published article

Journal ref: Mathematical Biosciences 403 (2027) 109844

DOI: 10.1016/j.mbs.2026.109844

Code: https://github.com/omaclaren/reparam

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 91/100

The gist: Invariant Image Reparameterisation (IIR) is an approach to parameter identifiability analysis that links symbolic reparameterisation ideas with local numerical calculations to determine a…

Key concepts

Invariant Image
This is a reduced, basis-independent mathematical representation of the parameter combinations that control how an observable model behaves. It captures the essential information about parameter dependencies regardless of the specific coordinate system used to describe them.
Symbolic Reparameterisation Condition
This is a theoretical condition derived from differential equations on the row and null spaces of an auxiliary mapping Jacobian. It describes when a transformation can be written symbolically, which is then approximated numerically at a single reference point.
Sparse Monomial Bases
These are specific coordinate representations chosen for the invariant image. They are often preferred over standard methods because they yield parameter combinations that are more interpretable and mechanistically meaningful for understanding the model's underlying physics or biology.

Terminology

Summary

Invariant Image Reparameterisation (IIR) is an approach to parameter identifiability analysis that links symbolic reparameterisation ideas with local numerical calculations to determine a basis-independent representation of parameter combinations associated with observable model behaviour. p. 0

Where the paper says it

Here, we consider Invariant Image Reparameterisation (IIR), which asks when symbolic reparameterisation conditions can be replaced by numerical derivative calculations at a single reference point. The central object is the invariant image: a reduced, basis-independent representation of the parameter combinations controlling observable model behaviour.

This method is significant because it allows the symbolic reparameterisation condition to be replaced by a single numerical Jacobian calculation at a single reference point, bridging symbolic and numerical methods for identifiability analysis, model reduction, and prediction. p. 0
Where the paper says it

Such issues motivate model reparameterisation and reduction methods. Here, we consider Invariant Image Reparameterisation (IIR), which asks when symbolic reparameterisation conditions can be replaced by numerical derivative calculations at a single reference point.

Core Concept of Invariant Image Reparameterisation

The central object in this approach is the invariant image, which is defined as a reduced, basis-independent representation of the parameter combinations controlling observable model behaviour. The authors show that when a one-to-one componentwise transformation makes observable behaviour depend only on fixed linear combinations of the transformed parameters, a single numerical Jacobian determines the associated lower-dimensional reparameterisation space. p. 0

Where the paper says it

The central object is the invariant image: a reduced, basis-independent representation of the parameter combinations controlling observable model behaviour. We show that when a one-toone componentwise transformation makes observable behaviour depend only on fixed linear combinations of the transformed parameters, a single numerical Jacobian determines the associated lower-dimensional reparameterisation space.

This invariant image admits multiple coordinate representations, such as the SVD which gives a default orthonormal basis ordered by local identifiability, and sparse monomial bases which are often more interpretable. p. 0
Where the paper says it

In structurally identifiable but practically weakly informed settings, the same calculations separate strongly and weakly informed parameter combinations. The invariant image admits multiple coordinate representations: the SVD gives a default orthonormal basis ordered by local identifiability, while sparse monomial bases are often more interpretable.

Derivation of Symbolic and Numerical Conditions

The symbolic reparameterisation condition is given by differential conditions on the row and null spaces of the auxiliary mapping Jacobian. For reparameterisations that can be written as a sequence involving a componentwise change of variables, a constant linear reduction, and an optional componentwise reparameterisation, this symbolic condition can be represented exactly by local derivative information at a single reference point. p. 2

Where the paper says it

First, we state an exact symbolic reparameterisation condition in terms of differential conditions on the row and null spaces of the auxiliary mapping Jacobian. Second, we show that, for reparameterisations that can be written as a sequence of transformations consisting of a componentwise change of variables, a constant linear reduction in the transformed variables, and an optional componentwise reparameterisation of the reduced coordinates, the symbolic condition can be represented exactly by local derivative information at a single reference point.

The key structural assumption is that the transformed auxiliary mapping can be written in terms of reduced transformed coordinates as phi∗(θ∗) = ϕ¯(Aθ∗) = ϕ¯(η). p. 6
Where the paper says it

The key structural assumption in our approach is that we can write the transformed auxiliary mapping in terms of reduced transformed coordinates: (2. 11) ϕ∗(θ ∗ ) = ϕ¯(Aθ∗ ) = ϕ¯(η).

This leads to the conditions null A ⊆ null Dθ∗ ϕ∗(θ∗), row A ⊇ row Dθ∗ ϕ∗(θ∗). p. 7
Where the paper says it

Hence, for this representation, we have the conditions: (2. 13) null A ⊆ null Dθ ∗ ϕ∗(θ ∗ ), row A ⊇ row Dθ ∗ ϕ∗(θ ∗ ).

Extraction of Invariant Subspaces

The first-order invariance condition is used to distinguish minimal from non-minimal reductions by finding the invariant part of the local null space. The calculation involves computing a stacked matrix C0 = 0, where the columns of C0 are coefficient vectors for those local null directions that satisfy the first-order invariance condition. p. 7

Where the paper says it

We later do this using a first-order invariance calculation. While the above condition does not uniquely determine A, as any basis for the relevant null space and complementary row space can be used, it establishes the existence of a matrix capturing the non-identifiable directions and the retained, potentially identifiable parameter combinations.

A default choice for the invariant null basis is N = V0VM0, and its canonical orthonormal complement is N⊥ = Vr V0VMr. p. 0
Where the paper says it

In structurally identifiable but practically weakly informed settings, the same calculations separate strongly and weakly informed parameter combinations. The invariant image admits multiple coordinate representations: the SVD gives a default orthonormal basis ordered by local identifiability, while sparse monomial bases are often more interpretable.

Practical Applications and Basis Choice

The method allows for the choice of coordinate representations within the invariant image, such as the default SVD construction or sparse monomial bases. Sparse monomial bases are often more interpretable than the SVD basis because they can provide simpler and more mechanistically meaningful parameter combinations while spanning the same invariant row or invariant null subspaces. p. 0

Where the paper says it

In structurally identifiable but practically weakly informed settings, the same calculations separate strongly and weakly informed parameter combinations. The invariant image admits multiple coordinate representations: the SVD gives a default orthonormal basis ordered by local identifiability, while sparse monomial bases are often more interpretable.

The Profile-Wise Analysis (PWA) framework uses these coordinates to provide likelihood-based uncertainty quantification and prediction. p. 0
Where the paper says it

The invariant image admits multiple coordinate representations: the SVD gives a default orthonormal basis ordered by local identifiability, while sparse monomial bases are often more interpretable. Treating these coordinates as interest parameters in Profile-Wise Analysis gives likelihood-based uncertainty quantification and prediction.

Model Examples and Results

The method is demonstrated on a parameterised normal model, where the Poisson-limit case reproduces the expected identifiable combination np and invariant null coordinate n/p. For the repressilator, the sparse monomial basis search recovers natural mechanistic ratios and products from an 18-parameter nonlinear ODE model. p. 4

Where the paper says it

The clearest class of models, and the class we mainly focus on here, is one in which the model can be expressed in terms of monomial parameter combinations. This is a natural assumption for many mechanistic models, for example via dimensional analysis [10, 2], and also arises in asymptotic approximations and model reduction [47].

The analysis shows that for observed mRNA trajectories, prediction uncertainty is mainly carried by the identifiable ratio K1/β1, while the non-identifiable product β1K1 has little effect on the observed outputs. p. 2
Where the paper says it

To illustrate these ideas, we first apply our method to a parameterised normal model with practical identifiability issues and a limiting structurally non-identifiable model, before considering a more complex mechanistic model of gene regulation, the repressilator [29, 27]. The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible.

The augmented IIR calculation shows that observing the first protein would remove this null direction while retaining those for genes 2 and 3. p. 9
Where the paper says it

MACLAREN ET AL. Θ Ψ Φ ϕ(θ) ϕ˜(ψ) Θ ∗ f(θ) g(η) = g(Af(θ)) ψ(θ) η = Aθ∗ = Af(θ) H Figure 1: The upper route shows the basic image decomposition ϕ(θ) = ϕ˜(ψ(θ)), while the lower route shows the transformed-coordinate construction used in IIR.

Conclusion and Future Directions

The paper concludes that IIR can replace symbolic global calculations with equivalent numerical calculations up to numerical tolerances, providing analogous information in settings with practical rather than structural identifiability issues. The authors suggest future work should focus on systematically extending the framework to more general sequential or multi-stage constructions and developing more systematic methods for choosing interpretable bases within the invariant image. p. 2

Where the paper says it

We also consider sparse monomial bases, which are typically more interpretable. Coordinates chosen within the invariant image support model reduction and provide natural parameters of interest for our Profile-Wise Analysis (PWA) framework [64, 63], yielding targeted interval estimates for both interest parameters and predictions.

The approach is also applicable to stochastic and simulation-based settings, as well as partial differential equations. p. 3
Where the paper says it

Although our main mechanistic example involves ordinary differential equations in time, identifiability questions for partial differential equation models are also important in mathematical biology [59, 8]. The auxiliary-map approach we use here requires only a numerically computable mapping from parameters to observable quantities, together with numerically computable derivatives of this mapping.

Supplementary Material Details

The supplementary material provides detailed technical foundations, including Fisher information analysis, the transformation law for observed Fisher information at the maximum likelihood estimate, and the numerical details of the sparse monomial basis selection algorithm. This includes a greedy selection procedure based on subspace filter and rank filter to select interpretable bases. p. 0

Where the paper says it

The invariant image admits multiple coordinate representations: the SVD gives a default orthonormal basis ordered by local identifiability, while sparse monomial bases are often more interpretable. Treating these coordinates as interest parameters in Profile-Wise Analysis gives likelihood-based uncertainty quantification and prediction.

The results for the repressilator example are saved in a specific file and postprocessed using repository scripts. p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

Author Contributions

All authors contributed to the review and editing of the manuscript, with OJM writing the original draft with contributions from MJS. OJM developed the method details and proofs, with assistance from RN, and wrote the final code drawing on PWA-related code. p. 3

Where the paper says it

In our examples, the required derivatives are evaluated by automatic differentiation in Julia [60]. This is distinct from symbolic differentiation: automatic differentiation at the source-code level evaluates the numerical value of the derivative at a fixed input value [33].

Both OJM and MJS contributed to the conceptualisation of the article, choice of examples and design of numerical experiments. p. 3
Where the paper says it

Hence, it can in principle be applied to numerical solutions of partial differential equations or boundary-value ordinary differential equations. An example of the latter is considered in the repository examples.

Competing Interests

The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions are available in the paper, supplementary material, and the GitHub repository. p. 2

Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

References

The references listed cover various aspects of identifiability analysis, model reduction, and related statistical methods. Key works include those on Fisher information, sensitivity-based methods, and structural identifiability analysis. p. 1

Where the paper says it

Profile likelihood methods [38, 58, 47, 63] provide powerful practical identifiability diagnostics but usually require the user to choose target parameters or subsets in advance, or to perform significant manual analysis to guide reduction [27, 47]. Sloppiness and related approaches [7, 34, 52] can highlight locally dominant and weak directions but are closely related to, and share similar limitations with, Fisher information and sensitivity-based methods, and often rely on non-parameterisation-invariant approximations for uncertainty quantification.

The references also include foundational texts on sets for mathematics and specific software tools for testing global identifiability of biological and physiological systems.

Data Availability

All data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The authors wish to acknowledge use of the eResearch Infrastructure Platform hosted by the Crown company, Research and Education Advanced Network New Zealand (REANNZ) Ltd. p. 0

Where the paper says it

Structural non-identifiability occurs when distinct parameter values generate precisely the same observable behaviour, leading to non-uniqueness in inversion; practical non-identifiability occurs when different parameter values generate outputs too similar to be reliably distinguished given the noise level and available outputs, leading to instability or lack of precision in inversion. Identifiability is- ∗Department of Engineering Science and Biomedical Engineering, University of Auckland, Auckland 1142, New Zealand (oliver.

, and funded by the Ministry of Business, Innovation & Employment. The MJS was supported by the Australian Research Council Centre of Excellence for the Mathematical Analysis of Cellular Systems grant no. p. 0
Where the paper says it

†School of Mathematical Sciences, Queensland University of Technology (QUT), Brisbane, Australia. ‡ARC Centre of Excellence for the Mathematical Analysis of Cellular Systems, QUT, Brisbane, Australia.

CE230100001 (MJS). All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

p. 2
Where the paper says it

The repressilator provides a simple case in which the model structure, without additional transformations, falls outside the standard polynomial/rational form required by some differentialalgebra approaches to symbolic identifiability analysis, while the model outputs and derivatives remain numerically accessible. Additional examples are available in the accompanying repository and summarised briefly in the Supplementary Material.

The URL is https://github. p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

p. 0
Where the paper says it

We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.

com/omaclaren/reparam. p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

p. 0
Where the paper says it

A Julia implementation of IIR, with these and further examples, is available at https://github. com/omaclaren/reparam.

All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests. The data and code needed to evaluate the conclusions in the paper are present in the paper, supplementary material, and the GitHub repository. The URL is https://github.com/omaclaren/reparam. All technical content, proofs, algorithms, analyses, and final wording were developed, checked, and approved by the authors. The authors declare that they have no competing interests <ref:2502. p. 0
Where the paper says it

‡ARC Centre of Excellence for the Mathematical Analysis of Cellular Systems, QUT, Brisbane, Australia. 1 arXiv:2502.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Automatic derivation of reparameterization conditions from local numerical calculations allows AI systems to replace complex symbolic calculation with a single-point Jacobian evaluation, making identifiability analysis accessible for models lacking explicit rational forms.

  2. The invariant image representation provides a basis independent of coordinate choices, enabling AI systems to use the SVD or sparse monomial bases for parameter combinations that are either ordered by local identifiability or more mechanistically interpretable.

  3. First-order invariance calculations can be used to extract constant invariant null directions from local numerical derivatives, allowing AI systems to determine whether a reduced parameterization is minimal or non-minimal purely through numerical computation.

  4. The framework supports sequential application of IIR, where the reduced coordinates obtained from one image reparameterization can serve as the starting point for further reductions, enabling AI systems to find more general structures like additive rather than monomial reductions.

  5. Profile-wise analysis, integrated with IIR, enables AI systems to quantify uncertainty using likelihood-based confidence sets that separate information concerning identifiable and non-identifiable parameter combinations into distinct coordinate spaces.

Abstract

Structural and practical parameter non-identifiability issues are common when mathematical models are used to interpret data. Such issues motivate model reparameterisation and reduction methods. Here, we consider Invariant Image Reparameterisation (IIR), which asks when symbolic reparameterisation conditions can be replaced by numerical derivative calculations at a single reference point. The central object is the invariant image: a reduced, basis-independent representation of the parameter combinations controlling observable model behaviour. We show that when a one-to-one componentwise transformation makes observable behaviour depend only on fixed linear combinations of the transformed parameters, a single numerical Jacobian determines the associated lower-dimensional reparameterisation space. This includes models depending on monomial combinations of the original parameters. We also give a first-order invariance condition that can rule out a minimal reduction of this form and extracts candidate invariant null directions for an exact non-minimal reduction. In structurally identifiable but practically weakly informed settings, the same calculations separate strongly and weakly informed parameter combinations. The invariant image admits multiple coordinate representations: the SVD gives a default orthonormal basis ordered by local identifiability, while sparse monomial bases are often more interpretable. Treating these coordinates as interest parameters in Profile-Wise Analysis gives likelihood-based uncertainty quantification and prediction. We demonstrate the method on parameterised normal models with Poisson-limit, extended Poisson-limit, and non-limit cases, and on the repressilator, a nonlinear differential equation model of gene regulation. A Julia implementation of IIR, with these and further examples, is available at https://github.com/omaclaren/reparam.

Sources

Related papers