Bounding Classical and Quantum Correlations in Bayesian Networks with Quasiprobabilities

summary

Video file (mp4)

The gist

The gist The authors formalize the set of quasidistributions consistent with a graph, generalizing results from Bell scenarios to show that this quasi set coincides with the nested Markov model for

In short

The authors investigate correlations in Bayesian networks using quasiprobabilities—distributions that allow for negative probabilities—to generalize results from Bell scenarios. They define a 'quasi set' based on these models and prove that for tree-structured correlation scenarios, this quasi set is identical to the nested Markov model, providing a link between classical graphical models and quantum non-classicality.

Key concepts

Quasiprobabilities
These are probability distributions used in Bayesian networks that permit negative probabilities. While standard probabilities must be between zero and one, quasiprobabilities relax this constraint. This relaxation is crucial because it allows these models to generate all possible non-signaling correlations, which is a key feature of quantum mechanics.
Quasi Set
The quasi set is the collection of observed distributions that can arise as marginals from a full quasidistribution consistent with the network structure. It represents the broader class of correlations achievable when allowing unobserved nodes to follow quasiprobability distributions, extending beyond standard classical probability.
Nested Markov Model
This is a specific type of correlation model for Bayesian networks that has a simpler form, particularly in tree structures. The paper shows that the quasi set—the broader set derived from quasiprobabilities—coincides with this nested Markov model for tree-structured scenarios, suggesting this simpler classical structure captures the essential correlations.
Tree-Structured Correlation Scenario
This refers to a specific type of Bayesian network where observed variables have no children and latent variables have exactly two children. This structural constraint simplifies the mathematical analysis significantly, allowing the authors to establish that the quasi set equals the nested Markov model for this class of graphs.

Terminology used across episodes

This episode discusses

The paper

Bounding Classical and Quantum Correlations in Bayesian Networks with Quasiprobabilities · Read on arXiv

University of Oxford

Bell's theorem reveals that quantum theory is in tension with classical causal reasoning and, in particular, the notion of local causality. This is now understood as a particular example of non-classicality in the study of correlations in (Bayesian) networks with both unobserved and observed nodes: the correlations are probability distributions over the observed nodes. There is a great deal of work aiming to understand the bounds on quantum and classical correlations in such networks and one approach is to consider outer approximations to the former. Along these lines, we consider quasiprobabilistic models for Bayesian networks, which can be seen as classical models but the probability distributions involving unobserved nodes are "replaced" with quasiprobabilities that respect normalisation but not positivity. We denote the set of correlations resulting from these models as the quasi set. Such models have a history in the study of Bell-type non-classicality where it has been shown that they can produce all non-signalling correlations. We show a generalisation of this result for a broad class of networks, which motivates a conjecture that the quasi set recovers the so called nested Markov model. Our work utilises a connection to tensor network decompositions, which may be of independent interest.

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: We're looking at the paper: "Bounding Classical and Quantum Correlations in Bayesian Networks with Quasiprobabilities".

Mira: Here's the gist of the paper. The gist The authors formalize the set of quasidistributions consistent with a graph, generalizing results from Bell scenarios to show that this quasi set coincides with the nested Markov model for tree-structured correlation scenarios.

Kai: Directed Graphical Models and Classical Sets Directed graphical models are a tool used frequently in all fields related to statistical and causal modeling, such as medicine

1: , psychology

2: , economics

3: , and machine learning

4: (Page 1).

Mira: In directed acyclic graph (DAG) modeling, each vertex v of a graph G is associated with a random variable Xv from a sample space Xv (Page 2).

Kai: A distribution consistent with a latent variable DAG must satisfy some independence constraints, but these are not sufficient.

Paper summary: Kai: So what does this paper actually claim about Bayesian networks and these quasiprobabilities? They're focusing on how to bound the correlations between variables in a network that has both observed and unobserved nodes >

Mira: The main thesis is that they are considering outer approximations to the classical bounds, and they use these quasiprobabilistic models to define a "quasi set" of correlations >

Lev: This quasi set is defined as the set of observed distributions which are marginals of full quasidistributions that factorize according to the graph structure >

Kai: They give three different ways to define what those full quasidistributions look like, they call them Definition one with quasilatents, Definition two with quasinoise variables, and Definition three defining the sets Wl(G), Wr(G), and Wn(G) > <ref:2606.23372#pg1>

Mira: The important part they establish is Theorem four which shows that these three different ways of looking at the models actually result in the exact same quasi set, meaning it doesn't matter which definition you use >

Lev: And then they make a conjecture, Conjecture one stating that this quasi set coincides with the nested Markov model for any directed acyclic graph G > <ref:2606.23372#pg1>

Kai: So what does this mean for us right now? It connects these relaxed models to a known classical concept, the nested Markov model, which is something we already use in causal modeling >

Mira: It's significant because it suggests that even when you allow for these non-classical probability behaviors with unobserved nodes, the resulting correlations are still constrained by this nested structure >

Lev: The paper also focuses on tree-structured correlation scenarios where observed variables have no children and latent variables have exactly two children, in those cases they prove that the quasi set is identical to the nested Markov model for that specific type of graph >

Kai: That proof uses tensor network decompositions, showing how a distribution in the ordinary Markov model can be written as a tensor decomposition with respect to an undirected graph Gt >

Mira: The paper shows this works by using Lemma thirty-four which connects the tensor decomposition to independence constraints, eventually proving that W(G) equals M(G) equals N(G) for tree-structured scenarios > <ref:2606.23372#pg1>

Lev: For real hardware, if we were running this on a system with these network constraints, the overhead would land somewhere in those tensor network calculations as they try to prove the linear independence of columns >

Conclusion: Kai: So wrapping up the paper "Bounding Classical and Quantum Correlations in Bayesian Networks with Quasiprobabilities", what's the big picture here? It’s about formalizing a new set of correlations that sits between classical probability and quantum reality >

Mira: The authors essentially show that by allowing for these quasiprobabilities, we can capture a much broader range of non-signaling correlations than standard classical models allow >

Lev: But the main result they highlight is that for specific network structures, like tree structures, this broad set of possibilities collapses back onto the nested Markov model >

Kai: So if you're driving or cooking right now and thinking about how information flows through a system, this paper suggests that even when you allow for some probabilistic fuzziness in the unobserved parts, the underlying structure remains constrained by these classical independence rules >

Mira: It moves the conversation from just finding quantum violations to understanding how these non-classical effects are organized within known graphical structures >

Lev: The implication is that if we want to build error correction for hardware based on these networks, we can use this nested Markov model as a tractable target because the quasi set seems to recover it in those cases >

Kai: So ultimately, they prove that for tree-structured correlation scenarios, the quasi set is exactly the nested Markov model, and that's what really matters here >

More episodes

← Home