Abstract:The problem of sampling according to the probability distribution minimizing a given free energy, using interacting particles unadjusted kinetic Langevin Monte Carlo, is addressed. In this setting, three sources of error arise, related to three parameters: the number of particles $N$, the discretization step size $h$, and the length of the trajectory $n$. The main result of the present work is a quantitative estimate of strong convergence in relative entropy, implying non-asymptotic bounds for the quadratic risk of Monte Carlo estimators for bounded observables. The numerical discretization scheme considered here is a second-order splitting method, as commonly used in practice. In addition to $N,h,n$, the dependency in the ambient dimension $d$ of the problem is also made explicit, under suitable conditions. The main results are proven under general conditions (regularity, moments, log-Sobolev inequality), for which tractable conditions are then provided. In particular, a Lyapunov analysis is conducted under more general conditions than previous works; the nonlinearity may not be small and it may not be convex along linear interpolations between measures.
What problem does this paper attempt to address?
### What problems does this paper attempt to solve?
This paper aims to solve the problem of sampling according to the probability distribution with minimized given free energy, especially using the non - adjusted Kinetic Langevin Monte Carlo method (KLMC). Specifically, the paper focuses on how to effectively control the error during the sampling process under the influence of the following three parameters:
1. **Number of particles \( N \)**: The number of interacting particle systems used to approximate the target distribution.
2. **Discretization step size \( h \)**: The step size of the numerical discretization scheme.
3. **Trajectory length \( n \)**: The physical trajectory length of the chain, that is, the number of chain transitions.
The specific background of these problems is as follows:
- **Problem description**: Consider an energy functional \( F: \mathcal{P}_2(X)\to\mathbb{R} \), where \( \mathcal{P}_2(X) \) is the set of probability measures with finite second moments. The associated free energy is:
\[
F(\mu)=F(\mu)+H(\mu)
\]
where \( H(\mu)=\int\log\mu\,d\mu \) is the Boltzmann entropy. Assume that \( F \) has a unique global minimum \( \bar{\mu}^* \) under certain conditions, satisfying the self - consistent equation:
\[
\bar{\mu}^*\propto\exp(-U_{\bar{\mu}^*}(x))\,dx
\]
where \( U_\mu(x)=\frac{\delta F}{\delta m}(\mu, x) \) is the linear derivative of \( F \) at \( \mu \).
- **Sampling problem**: In order to sample from \( \bar{\mu}^* \), a common method is to approximate its marginal equilibrium distribution. This usually involves simulating the system using the Markov chain Monte Carlo (MCMC) method. This paper specifically focuses on the Unadjusted Kinetic Langevin Chain, which is obtained by applying a splitting discretization scheme to the Kinetic Langevin diffusion.
- **Error sources**: Due to particle approximation, discretization error and long - time convergence to the stationary state, this method is affected by three types of errors. The main goal of the paper is to obtain non - asymptotic bounds of these errors under relative entropy, and explicitly depend on the environmental dimension \( d \) and other conditions.
### Main contributions
1. **Quantitative estimation**: The paper provides a quantitative estimate of strong convergence, which means that the non - asymptotic bound of the quadratic risk of the Monte Carlo estimator for bounded observations can be obtained.
2. **Generalize existing work**: Compared with previous work, this paper not only considers continuous - time processes, but also considers the numerical discretization schemes actually used. In addition, this paper relaxes the requirements for non - linear terms, allowing a larger non - linear range.
3. **Complexity analysis**: Discusses the time complexity and space complexity of the algorithm in different situations, especially its performance in high - dimensional problems.
Through these works, the paper provides a theoretical basis and technical means for understanding and optimizing the sampling method based on Kinetic Langevin diffusion.