-
A Systematization of the Wagner Framework: Graph Theory Conjectures and Reinforcement Learning
Authors:
Flora Angileri,
Giulia Lombardi,
Andrea Fois,
Renato Faraone,
Carlo Metta,
Michele Salvi,
Luigi Amedeo Bianchi,
Marco Fantozzi,
Silvia Giulia Galfrè,
Daniele Pavesi,
Maurizio Parton,
Francesco Morandin
Abstract:
In 2021, Adam Zsolt Wagner proposed an approach to disprove conjectures in graph theory using Reinforcement Learning (RL). Wagner's idea can be framed as follows: consider a conjecture, such as a certain quantity f(G) < 0 for every graph G; one can then play a single-player graph-building game, where at each turn the player decides whether to add an edge or not. The game ends when all edges have b…
▽ More
In 2021, Adam Zsolt Wagner proposed an approach to disprove conjectures in graph theory using Reinforcement Learning (RL). Wagner's idea can be framed as follows: consider a conjecture, such as a certain quantity f(G) < 0 for every graph G; one can then play a single-player graph-building game, where at each turn the player decides whether to add an edge or not. The game ends when all edges have been considered, resulting in a certain graph G_T, and f(G_T) is the final score of the game; RL is then used to maximize this score. This brilliant idea is as simple as innovative, and it lends itself to systematic generalization. Several different single-player graph-building games can be employed, along with various RL algorithms. Moreover, RL maximizes the cumulative reward, allowing for step-by-step rewards instead of a single final score, provided the final cumulative reward represents the quantity of interest f(G_T). In this paper, we discuss these and various other choices that can be significant in Wagner's framework. As a contribution to this systematization, we present four distinct single-player graph-building games. Each game employs both a step-by-step reward system and a single final score. We also propose a principled approach to select the most suitable neural network architecture for any given conjecture, and introduce a new dataset of graphs labeled with their Laplacian spectra. Furthermore, we provide a counterexample for a conjecture regarding the sum of the matching number and the spectral radius, which is simpler than the example provided in Wagner's original paper.
The games have been implemented as environments in the Gymnasium framework, and along with the dataset, are available as open-source supplementary materials.
△ Less
Submitted 18 June, 2024;
originally announced June 2024.
-
Limits of stochastic Volterra equations driven by Gaussian noise
Authors:
Luigi Amedeo Bianchi,
Stefano Bonaccorsi,
Martin Friesen
Abstract:
We study stochastic Volterra equations in Hilbert spaces driven by cylindrical Gaussian noise. We derive a mild formulation for the stochastic Volterra equation, prove the equivalence of mild and strong solutions, the existence and uniqueness of mild solutions, and study space-time regularity. Furthermore, we establish the stability of mild solutions in $L^q(\R_+)$, prove the existence of limit di…
▽ More
We study stochastic Volterra equations in Hilbert spaces driven by cylindrical Gaussian noise. We derive a mild formulation for the stochastic Volterra equation, prove the equivalence of mild and strong solutions, the existence and uniqueness of mild solutions, and study space-time regularity. Furthermore, we establish the stability of mild solutions in $L^q(\R_+)$, prove the existence of limit distributions in the Wasserstein $p$-distance with $p \in [1,\infty)$, and characterise when these limit distributions are independent of the initial state of the process despite the presence of memory. While our techniques allow for a general class of Volterra kernels, they are particularly suited for completely monotone kernels and fractional Riemann-Liouville kernels in the full range $α\in (0,2)$.
△ Less
Submitted 13 November, 2023;
originally announced November 2023.
-
Cosmoglobe DR1. III. First full-sky model of polarized synchrotron emission from all WMAP and Planck LFI data
Authors:
D. J. Watts,
U. Fuskeland,
R. Aurlien,
A. Basyrov,
L. A. Bianchi,
M. Brilenkov,
H. K. Eriksen,
K. S. F. Fornazier,
M. Galloway,
E. Gjerløw,
B. Hensley,
L. T. Hergt,
D. Herman,
H. Ihle,
K. Lee,
J. G. S. Lunde,
S. K. Nerval,
M. San,
N. O. Stutzer,
H. Thommesen,
I. K. Wehus
Abstract:
We present the first model of full-sky polarized synchrotron emission that is derived from all WMAP and Planck LFI frequency maps. The basis of this analysis is the set of end-to-end reprocessed Cosmoglobe Data Release 1 sky maps presented in a companion paper, which have significantly lower instrumental systematics than the legacy products from each experiment. We find that the resulting polarize…
▽ More
We present the first model of full-sky polarized synchrotron emission that is derived from all WMAP and Planck LFI frequency maps. The basis of this analysis is the set of end-to-end reprocessed Cosmoglobe Data Release 1 sky maps presented in a companion paper, which have significantly lower instrumental systematics than the legacy products from each experiment. We find that the resulting polarized synchrotron amplitude map has an average noise rms of $3.2\,\mathrm{μK}$ at 30 GHz and $2^{\circ}$ FWHM, which is 30% lower than the recently released BeyondPlanck model that included only LFI+WMAP Ka-V data, and 29% lower than the WMAP K-band map alone. The mean $B$-to-$E$ power spectrum ratio is $0.40\pm0.02$, with amplitudes consistent with those measured previously by Planck and QUIJOTE. Assuming a power law model for the synchrotron spectral energy distribution, and using the $T$--$T$ plot method, we find a full-sky inverse noise-variance weighted mean of $β_{\mathrm{s}}=-3.07\pm0.07$ between Cosmoglobe DR1 K-band and 30 GHz, in good agreement with previous estimates. In summary, the novel Cosmoglobe DR1 synchrotron model is both more sensitive and systematically cleaner than similar previous models, and it has a more complete error description that is defined by a set of Monte Carlo posterior samples. We believe that these products are preferable over previous Planck and WMAP products for all synchrotron-related scientific applications, including simulation, forecasting and component separation.
△ Less
Submitted 20 October, 2023;
originally announced October 2023.
-
Cosmoglobe DR1 results. I. Improved Wilkinson Microwave Anisotropy Probe maps through Bayesian end-to-end analysis
Authors:
D. J. Watts,
A. Basyrov,
J. R. Eskilt,
M. Galloway,
L. T. Hergt,
D. Herman,
H. T. Ihle,
S. Paradiso,
F. Rahman,
H. Thommesen,
R. Aurlien,
M. Bersanelli,
L. A. Bianchi,
M. Brilenkov,
L. P. L. Colombo,
H. K. Eriksen,
C. Franceschet,
U. Fuskeland,
E. Gjerløw,
B. Hensley,
G. A. Hoerning,
K. Lee,
J. G. S. Lunde,
A. Marins,
S. K. Nerval
, et al. (8 additional authors not shown)
Abstract:
We present Cosmoglobe Data Release 1, which implements the first joint analysis of WMAP and Planck LFI time-ordered data, processed within a single Bayesian end-to-end framework. This framework builds directly on a similar analysis of the LFI measurements by the BeyondPlanck collaboration, and approaches the CMB analysis challenge through Gibbs sampling of a global posterior distribution, simultan…
▽ More
We present Cosmoglobe Data Release 1, which implements the first joint analysis of WMAP and Planck LFI time-ordered data, processed within a single Bayesian end-to-end framework. This framework builds directly on a similar analysis of the LFI measurements by the BeyondPlanck collaboration, and approaches the CMB analysis challenge through Gibbs sampling of a global posterior distribution, simultaneously accounting for calibration, mapmaking, and component separation. The computational cost of producing one complete WMAP+LFI Gibbs sample is 812 CPU-hr, of which 603 CPU-hrs are spent on WMAP low-level processing; this demonstrates that end-to-end Bayesian analysis of the WMAP data is computationally feasible. We find that our WMAP posterior mean temperature sky maps and CMB temperature power spectrum are largely consistent with the official WMAP9 results. Perhaps the most notable difference is that our CMB dipole amplitude is $3366.2 \pm 1.4\ \mathrm{μK}$, which is $11\ \mathrm{μK}$ higher than the WMAP9 estimate and $2.5\ σ$ higher than BeyondPlanck; however, it is in perfect agreement with the HFI-dominated Planck PR4 result. In contrast, our WMAP polarization maps differ more notably from the WMAP9 results, and in general exhibit significantly lower large-scale residuals. We attribute this to a better constrained gain and transmission imbalance model. It is particularly noteworthy that the W-band polarization sky map, which was excluded from the official WMAP cosmological analysis, for the first time appears visually consistent with the V-band sky map. Similarly, the long standing discrepancy between the WMAP K-band and LFI 30 GHz maps is finally resolved, and the difference between the two maps appears consistent with instrumental noise at high Galactic latitudes. All maps and the associated code are made publicly available through the Cosmoglobe web page.
△ Less
Submitted 14 March, 2023;
originally announced March 2023.
-
A class of fractional Ornstein-Uhlenbeck processes mixed with a Gamma distribution
Authors:
Luigi Amedeo Bianchi,
Stefano Bonaccorsi,
Luciano Tubaro
Abstract:
We consider a sequence of fractional Ornstein-Uhlenbeck processes, that are defined as solutions of a family of stochastic Volterra equations with kernel given by the Riesz derivative kernel, and leading coefficients given by a sequence of independent Gamma random variables. We construct a new process by taking the empirical mean of this sequence. In our framework, the processes involved are not M…
▽ More
We consider a sequence of fractional Ornstein-Uhlenbeck processes, that are defined as solutions of a family of stochastic Volterra equations with kernel given by the Riesz derivative kernel, and leading coefficients given by a sequence of independent Gamma random variables. We construct a new process by taking the empirical mean of this sequence. In our framework, the processes involved are not Markovian, hence the analysis of their asymptotic behaviour requires some ad hoc construction. In our main result, we prove the almost sure convergence in the space of trajectories of the empirical means to a given Gaussian process, which we characterize completely.
△ Less
Submitted 8 November, 2022; v1 submitted 13 May, 2021;
originally announced May 2021.
-
Linear Stochastic Dyadic model
Authors:
Luigi Amedeo Bianchi,
Francesco Morandin
Abstract:
We discuss a stochastic interacting particles' system connected to dyadic models of turbulence, defining suitable classes of solutions and proving their existence and uniqueness. We investigate the regularity of a particular family of solutions, called moderate, and we conclude with existence and uniqueness of invariant measures associated with such moderate solutions.
We discuss a stochastic interacting particles' system connected to dyadic models of turbulence, defining suitable classes of solutions and proving their existence and uniqueness. We investigate the regularity of a particular family of solutions, called moderate, and we conclude with existence and uniqueness of invariant measures associated with such moderate solutions.
△ Less
Submitted 27 October, 2020;
originally announced October 2020.
-
The impact of white noise on a supercritical bifurcation in the Swift-Hohenberg equation
Authors:
Luigi Amedeo Bianchi,
Dirk Blömker
Abstract:
We consider the impact of additive Gaussian white noise on a supercritical pitchfork bifurcation in an unbounded domain. As an example we focus on the stochastic Swift-Hohenberg equation with polynomial nonlinearity. Here we identify the order where small noise first impacts the bifurcation. Using an approximation via modulation equations, we provide a tool to analyse how the noise influences the…
▽ More
We consider the impact of additive Gaussian white noise on a supercritical pitchfork bifurcation in an unbounded domain. As an example we focus on the stochastic Swift-Hohenberg equation with polynomial nonlinearity. Here we identify the order where small noise first impacts the bifurcation. Using an approximation via modulation equations, we provide a tool to analyse how the noise influences the dynamics close to a change of stability.
△ Less
Submitted 1 October, 2020; v1 submitted 19 May, 2020;
originally announced May 2020.
-
Stochastic Navier-Stokes Equations and Related Models
Authors:
Luigi Amedeo Bianchi,
Franco Flandoli
Abstract:
Regularization by noise for certain classes of fluid dynamic equations, a theme dear to Giuseppe Da Prato (see G. Da Prato and A. Debussche, Ergodicity for the 3D stochastic Navier-Stokes equations, J. Math. Pures Appl., 2003), is reviewed focusing on 3D Navier-Stokes equations and dyadic models of turbulence.
Regularization by noise for certain classes of fluid dynamic equations, a theme dear to Giuseppe Da Prato (see G. Da Prato and A. Debussche, Ergodicity for the 3D stochastic Navier-Stokes equations, J. Math. Pures Appl., 2003), is reviewed focusing on 3D Navier-Stokes equations and dyadic models of turbulence.
△ Less
Submitted 2 February, 2020;
originally announced February 2020.
-
Modulation equation and SPDEs on unbounded domains
Authors:
Luigi Amedeo Bianchi,
Dirk Blömker,
Guido Schneider
Abstract:
We consider the approximation via modulation equations for nonlinear SPDEs on unbounded domains with additive space time white noise. Close to a bifurcation an infinite band of eigenvalues changes stability, and we study the impact of small space-time white noise on this bifurcation.
As a first example we study the stochastic Swift-Hohenberg equation on the whole real line. Here due to the weak…
▽ More
We consider the approximation via modulation equations for nonlinear SPDEs on unbounded domains with additive space time white noise. Close to a bifurcation an infinite band of eigenvalues changes stability, and we study the impact of small space-time white noise on this bifurcation.
As a first example we study the stochastic Swift-Hohenberg equation on the whole real line. Here due to the weak regularity of solutions the standard methods for modulation equations fail, and we need to develop new tools to treat the approximation.
As an additional result we sketch the proof for local existence and uniqueness of solutions for the stochastic Swift-Hohenberg and the complex Ginzburg Landau equations on the whole real line in weighted spaces that allow for unboundedness at infinity of solutions, which is natural for translation invariant noise like space-time white noise. Moreover we use energy estimates to show that solutions of the Ginzburg-Landau equation are Hölder continuous and have moments in those functions spaces. This gives just enough regularity to proceed with the error estimates of the approximation result.
△ Less
Submitted 23 September, 2019; v1 submitted 8 June, 2017;
originally announced June 2017.
-
Structure function and fractal dissipation for an intermittent inviscid dyadic model
Authors:
Luigi Amedeo Bianchi,
Francesco Morandin
Abstract:
We study a generalization of the original tree-indexed dyadic model by Katz and Pavlović for the turbulent energy cascade of three-dimensional Euler equation. We allow the coefficients to vary with some restrictions, thus giving the model a realistic spatial intermittency. By introducing a forcing term on the first component, the fixed point of the dynamics is well defined and some explicit comput…
▽ More
We study a generalization of the original tree-indexed dyadic model by Katz and Pavlović for the turbulent energy cascade of three-dimensional Euler equation. We allow the coefficients to vary with some restrictions, thus giving the model a realistic spatial intermittency. By introducing a forcing term on the first component, the fixed point of the dynamics is well defined and some explicit computations allow to prove the rich multifractal structure of the solution. In particular the exponent of the structure function is concave in accordance with other theoretical and experimental models. Moreover anomalous energy dissipation happens in a fractal set of dimension strictly less than 3.
△ Less
Submitted 15 June, 2017; v1 submitted 5 July, 2016;
originally announced July 2016.
-
Additive noise destroys the random attractor close to bifurcation
Authors:
Luigi Amedeo Bianchi,
Dirk Blömker,
Meihua Yang
Abstract:
We provide an example for stabilization by noise. Our approach does not rely on monotonicity arguments due to the presence of higher order differential operators or mixing properties of the system as the noise might be highly degenerate. In the examples a scalar additive noise destroys a high-dimensional random attractor of a PDE on an unbounded domain. In the presence of small noise close to bifu…
▽ More
We provide an example for stabilization by noise. Our approach does not rely on monotonicity arguments due to the presence of higher order differential operators or mixing properties of the system as the noise might be highly degenerate. In the examples a scalar additive noise destroys a high-dimensional random attractor of a PDE on an unbounded domain. In the presence of small noise close to bifurcation all trajectories converge to a single stationary solution.
△ Less
Submitted 17 March, 2016;
originally announced March 2016.
-
Pattern size in Gaussian fields from spinodal decomposition
Authors:
Luigi Amedeo Bianchi,
Dirk Blömker,
Philipp Wacker
Abstract:
We study the two-dimensional snake-like pattern that arises in phase separation of alloys described by spinodal decomposition in the Cahn-Hilliard model. These are somewhat universal patterns due to an overlay of eigenfunctions of the Laplacian with a similar wave-number. Similar structures appear in other models like reaction-diffusion systems describing animal coats' patterns or vegetation patte…
▽ More
We study the two-dimensional snake-like pattern that arises in phase separation of alloys described by spinodal decomposition in the Cahn-Hilliard model. These are somewhat universal patterns due to an overlay of eigenfunctions of the Laplacian with a similar wave-number. Similar structures appear in other models like reaction-diffusion systems describing animal coats' patterns or vegetation patterns in desertification.
Our main result studies random functions given by cosine Fourier series with independent Gaussian coefficients, that dominate the dynamics in the Cahn-Hilliard model. This is not a cosine process, as the sum is taken over domains in Fourier space that not only grow and scale with a parameter of order $1/\varepsilon$, but also move to infinity. Moreover, the model under consideration is neither stationary nor isotropic.
To study the pattern size of nodal domains we consider the density of zeros on any straight line through the spatial domain. Using a theorem by Edelman and Kostlan and weighted ergodic theorems that ensure the convergence of the moving sums, we show that the average distance of zeros is asymptotically of order $\varepsilon$ with a precisely given constant.
△ Less
Submitted 17 November, 2015; v1 submitted 7 October, 2015;
originally announced October 2015.
-
Modulation Equation for SPDEs in unbounded domains with space-time white noise -- Linear Theory
Authors:
Luigi Amedeo Bianchi,
Dirk Blömker
Abstract:
We study the approximation of SPDEs on the whole real line near a change of stability via modulation or amplitude equations, which acts as a replacement for the lack of random invariant manifolds on extended domains. Due to the unboundedness of the underlying domain a whole band of infinitely many eigenfunctions changes stability. Thus we expect not only a slow motion in time, but also a slow spat…
▽ More
We study the approximation of SPDEs on the whole real line near a change of stability via modulation or amplitude equations, which acts as a replacement for the lack of random invariant manifolds on extended domains. Due to the unboundedness of the underlying domain a whole band of infinitely many eigenfunctions changes stability. Thus we expect not only a slow motion in time, but also a slow spatial modulation of the dominant modes, which is described by the modulation equation.
As a first step towards a full theory of modulation equations for nonlinear SPDEs on unbounded domains, we focus, in the results presented here, on the linear theory for one particular example, the Swift-Hohenberg equation. These linear results are one of the key technical tools to carry over the deterministic approximation results to the stochastic case with additive forcing. One technical problem for establishing error estimates rises from the spatially translation invariant nature of space-time white noise on unbounded domains, which implies that at any time we can expect the error to be always very large somewhere in space.
△ Less
Submitted 26 April, 2016; v1 submitted 22 July, 2015;
originally announced July 2015.
-
Uniqueness for an inviscid stochastic dyadic model on a tree
Authors:
Luigi Amedeo Bianchi
Abstract:
In this paper we prove that the lack of uniqueness for solutions of the tree dyadic model of turbulence is overcome with the introduction of a suitable noise. The uniqueness is a weak probabilistic uniqueness for all $l^2$-initial conditions and is proven using a technique relying on the properties of the $q$-matrix associated to a continuous time Markov chain.
In this paper we prove that the lack of uniqueness for solutions of the tree dyadic model of turbulence is overcome with the introduction of a suitable noise. The uniqueness is a weak probabilistic uniqueness for all $l^2$-initial conditions and is proven using a technique relying on the properties of the $q$-matrix associated to a continuous time Markov chain.
△ Less
Submitted 30 January, 2013;
originally announced January 2013.
-
A dyadic model on a tree
Authors:
David Barbato,
Luigi Amedeo Bianchi,
Franco Flandoli,
Francesco Morandin
Abstract:
We study an infinite system of non-linear differential equations coupled in a tree-like structure. This system was previously introduced in the literature and it is the model from which the dyadic shell model of turbulence was derived. It mimics 3d Euler and Navier-Stokes equations in a rough approximation of a wavelet decomposition. We prove existence of finite energy solutions, anomalous dissipa…
▽ More
We study an infinite system of non-linear differential equations coupled in a tree-like structure. This system was previously introduced in the literature and it is the model from which the dyadic shell model of turbulence was derived. It mimics 3d Euler and Navier-Stokes equations in a rough approximation of a wavelet decomposition. We prove existence of finite energy solutions, anomalous dissipation in the inviscid unforced case, existence and uniqueness of stationary solutions (either conservative or not) in the forced case.
△ Less
Submitted 12 July, 2012;
originally announced July 2012.
