- Rating
- 4(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes to learn distribution dynamics of autonomous dynamical systems (deterministic or stochastic). Conceptually, it follows the idea of Distributional Koopman Operator (Oprea et al., 2025), but adds deep components to learn statistics of empirical distribution and to reconstruct distribution from them. Practically, the paper proposes a conventional Encoder - Finite Koopman Evolution - Decoder model, but adapted in dynamic of distributions: the encoder is implemented by a permutation-invariant network and takes empirical samples as input; the decoder is implemented by conditional flow matching. Evaluated on different wasserstein distances, the experimental results show that the proposed DisKO learn distribution dynamics better than baselines. No theoretical results are extablished to support the performance of DisKO. Overall, the paper nicely combines distributional koopman operator and conditional flow matching, and proposes a new conceptual framework to learn distribution dynamics.
Critical strengths
- DisKO consists of a new conceptual framework to learn distribution dynamics and it nicely combines distributional koopman operator and conditional flow matching.
- Compared to transfer operator which is heavy to approximate, DisKO use a more compact representation to learn distribution dynamics
Critical weaknesses
- No theoretical result is established to support experimental results: it's hard to know whether the claimed distributional koopman operator is actually learned or well approximated
Questions for authors
- The Encoder - Finite Koopman evolution - Decoder pipeline approximates the transfer operator, why the comparison to the transfer operator is not discussed at all?
- Can the authors provide insights on how theoretical guarantee can be established?
5 comments
Response to Q2 (Theoretical guarantees, 1/3) — I. (a) Existence of approximately invariant subspaces and (b) Encoder approximation of distributional observables
Thank you for asking what theoretical guarantees DisKO admits. We give two results. (I) With the encoder fixed and under the assumptions of Proposition 1, the propagator fitted by least squares to exact snapshot pairs converges as to the optimal projection of the true distributional Koopman operator onto the chosen observable space. (II) We give conditions under which future distributions are identifiable in a class of linear Gaussian systems.
I. Finite-dimensional approximation of the distributional Koopman operator and approximately invariant subspaces
TL;DR. Closure relations in state space carry over to distribution space, where nonlinear observables can yield additional finite-dimensional closures (a). Existing results show that, under suitable compactness assumptions, a sufficiently expressive DisKO encoder can uniformly approximate any finite collection of continuous distributional observables to arbitrary accuracy (b). With the encoder fixed, we fit the dynamics to snapshot pairs. If the starting distributions represent the reference measure and the observables are nondegenerate, the least-squares fit converges to the optimal Koopman projection on this space as (c). This motivates DisKO's three stages. We learn observables by reconstruction, fix them to fit the dynamics, then jointly optimize all components to reduce the closure residual while retaining distributional information (d).
(a) Existence of approximately invariant subspaces
If state observables satisfy an exact or approximate closure relation under the underlying Markov generator, their expectations satisfy the same evolution equations with the same bound on the closure error. Nonlinear combinations of moments can yield additional closure relations for the same system.
Let be the generator of the underlying Markov dynamics. Suppose state observables satisfy
Here bounds the closure error at the generator level, with exact closure when . Define the distributional observables by . The distributional Koopman generator satisfies . Taking expectations therefore gives
where . Thus the distributional observables inherit the closure relation and its error bound.
Nonlinear distributional observables can have closed equations of their own. Consider the discrete-time doubling map on . For a single particle, evolves into after one step and after two. Each step introduces a new frequency. More generally, the doubling map is mixing with respect to Lebesgue measure. Its Koopman operator on is an isometry whose only eigenfunctions are constants. Its finite-dimensional invariant subspaces therefore contain only constant functions.
Now consider the same map acting on the family of distributions with densities , where . At the density is uniform. Larger values of place more mass toward the right end of the interval. A particle at after one step can come from or , so the new density is . Each step halves the slope of the density and keeps it in the same family.
Let , the variance of the uniform distribution minus the variance of the current particle distribution. We can estimate from the sample variance of a single snapshot. Integrating against gives
On this family, the expectation of any fixed state observable , , is affine in . No such expectation can therefore represent , which is quadratic in . The nonlinear observable spans an exact one-dimensional distributional Koopman invariant subspace. Since , its value uniquely determines and hence the full density after any number of steps.
(b) Encoder approximation of distributional observables
DisKO averages learned sample features before a neural readout (Zaheer et al., 2017). On a weakly compact family of distributions over a compact state space, such encoders with sufficient capacity can uniformly approximate any finite set of continuous observables to arbitrary accuracy (Pevný & Kovařík, 2019).
Response to Q2 (Theoretical guarantees, 2/3) — I. (c) Convergence to the optimal Koopman projection from finite snapshots and (d) How the theory informs three-stage training
(c) Convergence to the optimal Koopman projection from finite snapshots
Given a set of observables, such as those defined by a fixed encoder, we show when fitting the dynamics from snapshots converges to the optimal Koopman projection. The one-step propagation residual provides a way to assess approximate invariance of their span.
Fix the encoder . Write for its observable space and for the augmented observable vector. Let be a reference probability measure on the family of distributions . It determines how starting distributions are weighted in the population prediction error. Suppose we observe pairs of exact distribution snapshots at a fixed lag . The starting distributions define the empirical measure . Averaging with respect to gives each pair weight . The matrix minimizes the mean squared one-step error.
Thus represents the optimal finite-dimensional projection of the true Koopman action on under , extending the usual Koopman projection to distribution space.
Two sampling issues arise in practice. First, a finite snapshot gives an empirical distribution. For independent, identically distributed samples, the law of large numbers ensures that its encoding under a fixed encoder converges to that of the underlying distribution as the sample size grows. Second, even with exact snapshots, the propagation rule we can identify depends on which starting distributions we observe. Proposition 1 extends the classical projection consistency result for extended dynamic mode decomposition (EDMD) to this setting.
Proposition 1 (Consistency of Empirical Distributional Koopman Projection). Assume that is compact, and are continuous, and the following conditions hold.
- The empirical measure of the starting distributions converges weakly to the reference measure, .
- The Gram matrix is positive definite.
Then, as ,
M_\Pi^\ast=\mathbb E_\Pi[h(T_\Delta\nu)h(\nu)^\top]G_\Pi^{-1}.$$ Here $M_\Pi^\ast$ represents the $L^2(\Pi)$-orthogonal projection of the true distributional Koopman action on $V_\theta$ back onto $V_\theta$. **Remark (What the Gram condition requires of the snapshots).** More samples within a snapshot give a more accurate estimate of the distribution at that time. Identifying the propagator also requires enough variation across snapshots to distinguish different propagation rules. If all snapshots come from the same stationary distribution $\nu_\star$, even exact observations give the same vector $v=h(\nu_\star)$ each time. Then $G_\Pi=vv^\top$ has rank one, and every matrix satisfying $Mv=v$ fits the data. These snapshots cannot tell us whether a perturbation away from the stationary distribution would persist or decay. Thus, when the starting distributions represent the reference measure and the observables are nondegenerate, the fit from finite snapshots consistently estimates the optimal finite-dimensional projection of the true Koopman operator. #### (d) How the theory informs three-stage training Extending classical Koopman learning theory to distribution space raises questions about the existence of invariant subspaces and gives the Gram condition a concrete interpretation in terms of variation across distributions. **These results motivate DisKO's three-stage training procedure.** During **representation pretraining** in Stage 1, the distribution reconstruction objective encourages the encoder to learn observables that retain enough information to reconstruct a snapshot. In Stage 2, we fix the encoder and fit Koopman dynamics in the observable space it defines. For the discrete model at a fixed lag, suppose the ridge coefficient in Eq. (25) of the original paper is held fixed and the conditions of Proposition 1 hold. As the number $n$ of snapshot pairs tends to infinity, the fitted propagator converges to the matrix representing the optimal projection of the distributional Koopman action on this space under the reference measure. In Stage 3, **joint training** updates the observables, Koopman dynamics, and decoder together. The latent consistency loss reduces the Koopman closure residual, while the reconstruction and distribution prediction losses encourage the observables to retain the information needed for both tasks.
Response to Q1 (Transfer operators, 1/1) — DisKO and methods for learning the forward transfer operator
Thank you for pointing out this connection. For time-homogeneous Markov dynamics, DisKO's end-to-end predictor D(Ft(E(μ))) approximates the action of the forward transfer operator, Ptμ, on the family of distributions we model. Our original discussion focused on how distributional observables evolve under the Koopman operator (Oprea et al., 2025) and left this connection implicit.
Karimi and Georgiou (2022) use the Wasserstein distance to fit a common state map whose pushforward of each observed distribution approximates the next snapshot. Beier et al. (2025) estimate a transition kernel from batches of particles observed at two times, with the individual pairings within each batch unknown. For spectral analysis, BSO (Xu et al., 2026) learns the backward operator dual to the forward Markov operator. DisKO learns Ft(E(μ))≈E(Ptμ) from marginal distributions sampled separately at each time, without matching individuals across time.
WOT-iPSC measures different cells on different days. We compared DisKO directly with dynamic distribution decomposition (DDD), which can also learn from these snapshots (Taylor-King et al., 2020). DDD uses a continuous-time Markov rate matrix to evolve the mixture weights of Gaussian basis densities. We trained both methods on the 33 snapshots from days 0 to 15. Each forecast was a single rollout through day 18 from the same 1,024 cells observed on day 0. In this run (seed 0), DisKO had lower W1 at all six held-out time points, with a mean W1 32.4% lower than DDD's (Table 1.1).
Table 1.1. Extrapolation error at the six held-out time points on WOT-iPSC. We fit the SVD using only the training data. At each time point, we computed W1 by optimal assignment between 512 predicted and 512 observed samples in the 50-dimensional SVD coordinates on their original scale. The final row gives the equally weighted mean over the six time points.
References