Latent Dynamics · 2026-11-06
Identifying Neural Dynamics Using Nonlinear Interventional State Space Models
Hanna M. Tolle, Carles Balsells-Rodas, Yixin Wang, Pedro A. M. Mediano, Amin Nejatbakhsh
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
The authors propose a nonlinear interventional state-space model (iSSM) for identifying latent neural dynamics from high-dimensional observations under do-style interventions on latent variables. They prove affine identifiability for general injective emissions with Gaussian latent transitions, and permutation-scale-shift identifiability under interventions for general nonlinear latent dynamics. They validate on simulated FitzHugh–Nagumo and apply to decision-model adjudication and microstimulation prediction in macaque and human neural recordings.
Why it matters
Prior iSSM identifiability was restricted to linear latent dynamics and piecewise-linear emissions. This work extends those guarantees to much broader nonlinear latent dynamics and general injective emissions, enabling identification and prediction of responses to intervention combinations that were not seen during training, including proof-of-concept generalization to a held-out stimulation site in human data.
Method
- Model: iSSM with injective emission fθ, nonlinear latent transition gφ, observed exogenous intervention inputs u_t, and hard latent interventions via a binary mask m_t applied to the transition: x_{t+1}=m_t⊙gφ(x_t)+Bu_t+ε_t.
- Theory: show bounded completeness lets recovery of noiseless observations from noisy observations; then prove affine identifiability under injective emissions and Gaussian latent transitions; use interventions plus an unordered-pairs condition to reduce affine ambiguity to permutation, scaling, and shift.
- Evaluation: reconstruction, rollout accuracy, and latent accuracy (when ground-truth latents are available); apply to simulated decision-making models and to microstimulation response prediction in macaque and human iEEG with a held-out-site test tier.
Limitation
The authors treat the virtual stimulation analysis as a pilot: “Please note that we treat this as a pilot only. Input columns for many virtually stimulated sites are far from any sites actually stimulated in the dataset … and outside the scope of our identifiability results … Also, the present analysis uses only one subject and without independent replication.”
Abstract (from arXiv)
State-space models (SSMs) recover low-dimensional latent dynamics from high-dimensional observations, but latents are generally not identifiable from observational data alone. Interventional state-space models (iSSMs) use perturbations to break observational symmetries and identify parameters up to permutation, scaling, and shift. However, prior work restricts identifiability to linear latent dynamics and piecewise-linear emissions, excluding many behaviors characteristic of neural circuits. We extend iSSMs to a substantially broader model class: we prove that (i) affine identifiability holds under general injective emission functions and Gaussian latent transitions, and (ii) with interventions, identifiability up to permutation, scaling, and shift is preserved under general nonlinear latent dynamics. We validate our theory on simulated FitzHugh-Nagumo dynamics, and then demonstrate two applications: adjudicating between competing models of decision-making; and predicting responses to unseen micro-stimulation in macaque and human neural recordings, with proof-of-concept generalization to a held-out stimulation site in the latter. Finally, simulations yield practical guidance for designing stimulation protocols that improve identifiability.
Related papers
- Identifiable Signed Causal Learning via Latent Stochastic Differential Equations for Inferring Brain Dynamic Effective Connectivity
- When Latent Dynamics Meet Data Assimilation: A Benchmark of Koopman-Type Representations
- Identifiable World Models from Pretrained Diffusion Representations
- Learning Transferable Policies from Action-free Time Series Through Dynamical Embeddings
- Modeling Latent Disturbances for Robust Decision-Making in World Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-11-05 · Indexed by Pub-AI: 2026-11-06 UTC
Identifying Neural Dynamics Using Nonlinear Interventional State Space Models
Preprint posted: 2026-11-05 · Original record ↗
- arxiv
- 2611.08113
Explicit source versions (1)
- v1 · 2026-11-05
- Preprint available
Source date: 2026-11-05 · First detected: 2026-11-06 UTC · Historical baseline
Source evidence (3)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.08113v1; arxiv:announce_type=new; item pubDate (official metadata)
- Official evidence ↗ · atom:published (official metadata)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
State-space models (SSMs) recover low-dimensional latent dynamics from high-dimensional observations, but latents are generally not identifiable from observational data alone. Interventional state-space models (iSSMs) use perturbations to break observational symmetries and identify parameters up to permutation, scaling, and shift. However, prior work restricts identifiability to linear latent dynamics and piecewise-linear emissions, excluding many behaviors characteristic of neural circuits. We extend iSSMs to a substantially broader model class: we prove that (i) affine identifiability holds under general injective emission functions and Gaussian latent transitions, and (ii) with interventions, identifiability up to permutation, scaling, and shift is preserved under general nonlinear latent dynamics. We validate our theory on simulated FitzHugh-Nagumo dynamics, and then demonstrate two applications: adjudicating between competing models of decision-making; and predicting responses to unseen micro-stimulation in macaque and human neural recordings, with proof-of-concept generalization to a held-out stimulation site in the latter. Finally, simulations yield practical guidance for designing stimulation protocols that improve identifiability.
