Abstract:Solving decision problems in complex, stochastic environments is often achieved by estimating the expected outcome of decisions via Monte Carlo sampling. However, sampling may overlook rare, but important events, which can severely impact the decision making process. We present a method in which a Normalizing Flow generative model is trained to simulate samples directly from a conditional distribution given that a rare event occurs. By utilizing Coupling Flows, our model can, in principle, approximate any sampling distribution arbitrarily well. By combining the approximation method with Importance Sampling, highly accurate estimates of complicated integrals and expectations can be obtained. We include several examples to demonstrate how the method can be used for efficient sampling and estimation, even in high-dimensional and rare-event settings. We illustrate that by simulating directly from a rare-event distribution significant insight can be gained into the way rare events happen.

What problem does this paper attempt to address?

The paper attempts to address the problem of how to efficiently estimate the impact of rare but important events (i.e., events with very low probability) on decision outcomes in complex, stochastic environments. Traditional methods such as Monte Carlo simulation are inefficient in dealing with such problems because they struggle to capture these rare events. The paper proposes a generative model approach based on Normalizing Flows, which trains the model to directly generate samples from the conditional distribution given the occurrence of rare events. This method, combined with importance sampling techniques, can efficiently and accurately estimate complex integrals and expectations, thereby better understanding the system's behavior when rare events occur. Specifically, the main contributions of the paper include: 1. **Proposing a new framework**: Utilizing Normalizing Flows generative models to learn the optimal sampling distribution and use it to estimate the probabilities of rare events. 2. **Addressing the limitations of traditional methods**: Traditional parametric models are often not flexible enough to capture all the characteristics of complex systems; this method can approximate any target distribution without relying on specific problem knowledge. 3. **Improving the efficiency of rare event simulation**: By directly generating samples from the rare event conditional distribution, it significantly enhances the understanding and analysis of rare event behavior. In summary, the paper aims to improve the efficiency and accuracy of handling rare events in complex, stochastic environments by introducing a new generative model approach, thereby improving the decision-making process.

A Flow-Based Generative Model for Rare-Event Simulation

Efficient Rare Event Sampling with Unsupervised Normalising Flows

Rare Event Probability Learning by Normalizing Flows

Stochastic Sampling from Deterministic Flow Models

Stochastic Normalizing Flows

Generative Assignment Flows for Representing and Learning Joint Distributions of Discrete Data

EventFlow: Forecasting Continuous-Time Event Data with Flow Matching

Flow Matching with Gaussian Process Priors for Probabilistic Time Series Forecasting

Flow Map Matching

Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models

Flow Matching for Scalable Simulation-Based Inference

Conditional Pseudo-Reversible Normalizing Flow for Surrogate Modeling in Quantifying Uncertainty Propagation

Stochastic Generative Flow Networks

A Deep Reinforcement Learning Approach to Rare Event Estimation

SurroFlow: A Flow-Based Surrogate Model for Parameter Space Exploration and Uncertainty Quantification

Unifying Predictions of Deterministic and Stochastic Physics in Mesh-reduced Space with Sequential Flow Generative Model

MCFlow: Monte Carlo Flow Models for Data Imputation

Diffusion Generative Flow Samplers: Improving learning signals through partial trajectory optimization

A Tale of Two Latent Flows: Learning Latent Space Normalizing Flow with Short-run Langevin Flow for Approximate Inference

Generative Flow Networks: a Markov Chain Perspective

Uncertainty quantification and out-of-distribution detection using surjective normalizing flows