Saved in:
Bibliographic Details
Main Authors: Rozada, Sergio, Rey, Samuel, Mateos, Gonzalo, Marques, Antonio G.
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.21705
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915416849252352
author Rozada, Sergio
Rey, Samuel
Mateos, Gonzalo
Marques, Antonio G.
author_facet Rozada, Sergio
Rey, Samuel
Mateos, Gonzalo
Marques, Antonio G.
contents Dynamic programming (DP) is a fundamental tool used across many engineering fields. The main goal of DP is to solve Bellman's optimality equations for a given Markov decision process (MDP). Standard methods like policy iteration exploit the fixed-point nature of these equations to solve them iteratively. However, these algorithms can be computationally expensive when the state-action space is large or when the problem involves long-term dependencies. Here we propose a new approach that unrolls and truncates policy iterations into a learnable parametric model dubbed BellNet, which we train to minimize the so-termed Bellman error from random value function initializations. Viewing the transition probability matrix of the MDP as the adjacency of a weighted directed graph, we draw insights from graph signal processing to interpret (and compactly re-parameterize) BellNet as a cascade of nonlinear graph filters. This fresh look facilitates a concise, transferable, and unifying representation of policy and value iteration, with an explicit handle on complexity during inference. Preliminary experiments conducted in a grid-like environment demonstrate that BellNet can effectively approximate optimal policies in a fraction of the iterations required by classical methods.
format Preprint
id arxiv_https___arxiv_org_abs_2507_21705
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Unrolling Dynamic Programming via Graph Filters
Rozada, Sergio
Rey, Samuel
Mateos, Gonzalo
Marques, Antonio G.
Artificial Intelligence
Dynamic programming (DP) is a fundamental tool used across many engineering fields. The main goal of DP is to solve Bellman's optimality equations for a given Markov decision process (MDP). Standard methods like policy iteration exploit the fixed-point nature of these equations to solve them iteratively. However, these algorithms can be computationally expensive when the state-action space is large or when the problem involves long-term dependencies. Here we propose a new approach that unrolls and truncates policy iterations into a learnable parametric model dubbed BellNet, which we train to minimize the so-termed Bellman error from random value function initializations. Viewing the transition probability matrix of the MDP as the adjacency of a weighted directed graph, we draw insights from graph signal processing to interpret (and compactly re-parameterize) BellNet as a cascade of nonlinear graph filters. This fresh look facilitates a concise, transferable, and unifying representation of policy and value iteration, with an explicit handle on complexity during inference. Preliminary experiments conducted in a grid-like environment demonstrate that BellNet can effectively approximate optimal policies in a fraction of the iterations required by classical methods.
title Unrolling Dynamic Programming via Graph Filters
topic Artificial Intelligence
url https://arxiv.org/abs/2507.21705