Skip to content

Source code for "Model-Based Validation as Probabilistic Inference" presented at L4DC 2023

Notifications You must be signed in to change notification settings

sisl/ModelBasedValidationInference

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Model-based Validation as Probabilistic Inference

This is the codebase for the paper "Model-based Validation as Probabilistic Inference" presented at L4DC 2023.

We frame estimating the distribution over failure trajectories for sequential systems as Bayesian inference. Our model-based approach represents the distribution over failure trajectories using rollouts of system dynamics and computes trajectory gradients using automatic differentiation.

Our method efficiently samples from high-dimensional, multimodal failure distributions using an off-the-shelf implementation of Hamiltonian Monte Carlo. In the example below, we sample failures of a neural network-controlled inverted pendulum with noisy actions. The system fails when the pendulum falls over.

See our paper for more details and results! (Link coming soon)

Installation

First, clone this repo

$ git clone https://github.com/sisl/ModelBasedValidationInference.git

Enter the Julia REPL

$ julia

Type "]" to enter Pkg mode

julia> ]

Activate and instantiate the project

pkg> activate .
pkg> instantiate

Run one of the scripts from the scripts directory

$ julia --project scripts/run_pendulum.jl

About

Source code for "Model-Based Validation as Probabilistic Inference" presented at L4DC 2023

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages