Abstract:This note is concerned with weakly interacting stochastic particle systems with possibly singular pairwise interactions. In this setting, we observe a connection between entropic propagation of chaos and exponential concentration bounds for the empirical measure of the system. In particular, we establish a variational upper bound for the probability of a certain rare event, and then use this upper bound to show that "controlled" entropic propagation of chaos implies an exponential concentration bound for the empirical measure. This connection allows us to infer concentration bounds for a class of singular stochastic systems through a simple adaptation of the arguments developed in Jabin and Wang (2018).
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to establish an exponential concentration inequality of empirical measures in a weakly interacting stochastic particle system with a singular kernel. Specifically, the author focuses on how to derive the concentration bound of empirical measures by controlling the entropic propagation of chaos. This involves establishing strict mathematical estimates on the difference between the empirical measure of the system and the mean - field limit, especially when the kernel function \(K\) may be singular.
### Background of the paper
In many practical applications, such as fluid mechanics and biophysics, it is often necessary to deal with stochastic particle systems with singular kernels. There may be strong interactions among particles in these systems, resulting in the kernel function \(K\) being non - smooth or even discontinuous. In this case, traditional concentration inequality methods are often no longer applicable, so new mathematical tools and techniques are required to deal with such problems.
### Main contributions
1. **Connecting entropic propagation of chaos and concentration inequalities**: The author establishes the connection between entropic propagation of chaos and concentration inequalities of empirical measures. Specifically, they prove that if a system satisfies a certain form of "controlled" entropic propagation of chaos, an exponential concentration inequality of empirical measures can be derived.
2. **Generalizing existing results**: The author's work extends previous results on regular kernel functions, making them applicable to a wider range of singular kernel functions. This provides a theoretical basis for dealing with complex systems in practical applications.
3. **Technical details**: To achieve the above goals, the author introduces the "controlled Liouville equation" and, through a series of elaborate mathematical analyses, proves that under certain conditions, the solution of this equation can be used to derive the required concentration inequality.
### Mathematical expressions
- **Entropic propagation of chaos**:
\[
H_N(\rho^N_T\mid\rho^{\otimes N}_T)\leq H_N(\rho^N_0\mid\rho^{\otimes N}_0)+\frac{C}{N}
\]
where \(H_N\) represents the rescaled relative entropy.
- **Concentration inequality**:
\[
P\left[d_p(m^N_X,\rho_T)>\epsilon\right]\leq C_{\text{con},p}\exp\left(-C^{-1}_{\text{con},p}a_p(\epsilon)N\right)
\]
where \(d_p\) represents the \(p\)-Wasserstein distance, and the specific form of \(a_p(\epsilon)\) depends on \(p\) and \(d\).
### Conclusion
This paper provides a new method for dealing with stochastic particle systems with singular kernels by establishing the connection between entropic propagation of chaos and concentration inequalities. This result is not only of great theoretical significance but also provides a powerful tool for the analysis of complex systems in practical applications.