skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.


Title: Learning particle swarming models from data with Gaussian processes
Interacting particle or agent systems that exhibit diverse swarming behaviors are prevalent in science and engineering. Developing effective differential equation models to understand the connection between individual interaction rules and swarming is a fundamental and challenging goal. In this paper, we study the data-driven discovery of a second-order particle swarming model that describes the evolution of N N particles in R d \mathbb {R}^d under radial interactions. We propose a learning approach that models the latent radial interaction function as Gaussian processes, which can simultaneously fulfill two inference goals: one is the nonparametric inference of the interaction function with pointwise uncertainty quantification, and the other is the inference of unknown scalar parameters in the noncollective friction forces of the system. We formulate the learning problem as a statistical inverse learning problem and introduce an operator-theoretic framework that provides a detailed analysis of recoverability conditions, establishing that a coercivity condition is sufficient for recoverability. Given data collected from M M i.i.d trajectories with independent Gaussian observational noise, we provide a finite-sample analysis, showing that our posterior mean estimator converges in a Reproducing Kernel Hilbert Space norm, at an optimal rate in M M equal to the one in the classical 1-dimensional Kernel Ridge regression. As a byproduct, we show we can obtain a parametric learning rate in M M for the posterior marginal variance using L ∞<#comment/> L^{\infty } norm and that the rate could also involve N N and L L (the number of observation time instances for each trajectory) depending on the condition number of the inverse problem. We provide numerical results on systems exhibiting different swarming behaviors, highlighting the effectiveness of our approach in the scarce, noisy trajectory data regime.  more » « less
Award ID(s):
2111303 2340631
PAR ID:
10625969
Author(s) / Creator(s):
; ; ;
Publisher / Repository:
AMS
Date Published:
Journal Name:
Mathematics of Computation
Volume:
93
Issue:
349
ISSN:
0025-5718
Page Range / eLocation ID:
2391 to 2437
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. We formulate and prove a Conner–Floyd isomorphism for the algebraic K-theory of arbitrary qcqs derived schemes. To that end, we study a stable ∞<#comment/> \infty -category of non- A 1 \mathbb {A}^1 -invariant motivic spectra, which turns out to be equivalent to the ∞<#comment/> \infty -category of fundamental motivic spectra satisfying elementary blowup excision, previously introduced by the first and third authors. We prove that this ∞<#comment/> \infty -category satisfies P 1 \mathbb {P}^1 -homotopy invariance and weighted A 1 \mathbb {A}^1 -homotopy invariance, which we use in place of A 1 \mathbb {A}^1 -homotopy invariance to obtain analogues of several key results from A 1 \mathbb {A}^1 -homotopy theory. These allow us in particular to define a universal oriented motivic E ∞<#comment/> \mathbb {E}_\infty -ring spectrum M G L \mathrm {MGL} . We then prove that the algebraic K-theory of a qcqs derived scheme X X can be recovered from its M G L \mathrm {MGL} -cohomology via a Conner–Floyd isomorphism\[ M G L ∗<#comment/> ∗<#comment/> ( X ) ⊗<#comment/> L Z [ β<#comment/> ±<#comment/> 1 ] ≃<#comment/> K ∗<#comment/> ∗<#comment/> ( X ) , \mathrm {MGL}^{**}(X)\otimes _{\mathrm {L}{}}\mathbb {Z}[\beta ^{\pm 1}]\simeq \mathrm {K}{}^{**}(X), \]where L \mathrm {L}{} is the Lazard ring and K p , q ( X ) = K 2 q −<#comment/> p ( X ) \mathrm {K}{}^{p,q}(X)=\mathrm {K}{}_{2q-p}(X) . Finally, we prove a Snaith theorem for the periodized version of M G L \mathrm {MGL}
    more » « less
  2. A subset R R of integers is a set of Bohr recurrence if every rotation on T d \mathbb {T}^d returns arbitrarily close to zero under some non-zero multiple of R R . We show that the set { k ! 2 m 3 n :<#comment/> k , m , n ∈<#comment/> N } \{k!\, 2^m3^n\colon k,m,n\in \mathbb {N}\} is a set of Bohr recurrence. This is a particular case of a more general statement about images of such sets under any integer polynomial with zero constant term. We also show that if P P is a real polynomial with at least one non-constant irrational coefficient, then the set { P ( 2 m 3 n ) :<#comment/> m , n ∈<#comment/> N } \{P(2^m3^n)\colon m,n\in \mathbb {N}\} is dense in T \mathbb {T} , thus providing a joint generalization of two well-known results, one of Furstenberg and one of Weyl. 
    more » « less
  3. Let Ω<#comment/> + ⊂<#comment/> R n + 1 \Omega ^+\subset \mathbb {R}^{n+1} be a bounded δ<#comment/> \delta -Reifenberg flat domain, with δ<#comment/> > 0 \delta >0 small enough, possibly with locally infinite surface measure. Assume also that Ω<#comment/> −<#comment/> = R n + 1 ∖<#comment/> Ω<#comment/> + ¯<#comment/> \Omega ^-= \mathbb {R}^{n+1}\setminus \overline {\Omega ^+} is an NTA (non-tangentially accessible) domain as well and denote by ω<#comment/> + \omega ^+ and ω<#comment/> −<#comment/> \omega ^- the respective harmonic measures of Ω<#comment/> + \Omega ^+ and Ω<#comment/> −<#comment/> \Omega ^- with poles p ±<#comment/> ∈<#comment/> Ω<#comment/> ±<#comment/> p^\pm \in \Omega ^\pm . In this paper we show that the condition that log ⁡<#comment/> d ω<#comment/> −<#comment/> d ω<#comment/> + ∈<#comment/> VMO ⁡<#comment/> ( ω<#comment/> + ) \log \dfrac {d\omega ^-}{d\omega ^+} \in \operatorname {VMO}(\omega ^+) is equivalent to Ω<#comment/> + \Omega ^+ being a chord-arc domain with inner unit normal belonging to VMO ⁡<#comment/> ( H n | ∂<#comment/> Ω<#comment/> + ) \operatorname {VMO}(\mathcal {H}^n|_{\partial \Omega ^+})
    more » « less
  4. We show that for primes N , p ≥<#comment/> 5 N, p \geq 5 with N ≡<#comment/> −<#comment/> 1 mod p N \equiv -1 \bmod p , the class number of Q ( N 1 / p ) \mathbb {Q}(N^{1/p}) is divisible by p p . Our methods are via congruences between Eisenstein series and cusp forms. In particular, we show that when N ≡<#comment/> −<#comment/> 1 mod p N \equiv -1 \bmod p , there is always a cusp form of weight 2 2 and level Γ<#comment/> 0 ( N 2 ) \Gamma _0(N^2) whose ℓ<#comment/> \ell th Fourier coefficient is congruent to ℓ<#comment/> + 1 \ell + 1 modulo a prime above p p , for all primes ℓ<#comment/> \ell . We use the Galois representation of such a cusp form to explicitly construct an unramified degree- p p extension of Q ( N 1 / p ) \mathbb {Q}(N^{1/p})
    more » « less
  5. We prove and extend the longest-standing conjecture in ‘ q , t q,t -Catalan combinatorics,’ namely, the combinatorial formula for ∇<#comment/> m s μ<#comment/> \nabla ^m s_{\mu } conjectured by Loehr and Warrington, where s μ<#comment/> s_{\mu } is a Schur function and ∇<#comment/> \nabla is an eigenoperator on Macdonald polynomials. Our approach is to establish a stronger identity of infinite series of G L l GL_l characters involvingSchur Catalanimals; these were recently shown by the authors to represent Schur functions s μ<#comment/> [ −<#comment/> M X m , n ] s_{\mu }[-MX^{m,n}] in subalgebras Λ<#comment/> ( X m , n ) ⊂<#comment/> E \Lambda (X^{m,n})\subset \mathcal {E} isomorphic to the algebra of symmetric functions Λ<#comment/> \Lambda over Q ( q , t ) \mathbb {Q} (q,t) , where E \mathcal {E} is the elliptic Hall algebra of Burban and Schiffmann. We establish a combinatorial formula for Schur Catalanimals as weighted sums of LLT polynomials, with terms indexed by configurations of nested lattice paths callednests, having endpoints and bounding constraints controlled by data called aden. The special case for Λ<#comment/> ( X m , 1 ) \Lambda (X^{m,1}) proves the Loehr-Warrington conjecture, giving ∇<#comment/> m s μ<#comment/> \nabla ^m s_{\mu } as a weighted sum of LLT polynomials indexed by systems of nested Dyck paths. In general, for Λ<#comment/> ( X m , n ) \Lambda (X^{m,n}) our formula implies a new ( m , n ) (m,n) version of the Loehr-Warrington conjecture. In the case where each nest consists of a single lattice path, the nests in a den formula reduce to our previous shuffle theorem for paths under any line. Both this and the ( m , n ) (m,n) Loehr-Warrington formula generalize the ( k m , k n ) (km,kn) shuffle theorem proven by Carlsson and Mellit (for n = 1 n=1 ) and Mellit. Our formula here unifies these two generalizations. 
    more » « less