Quantum Bayesian Networks: Compositionality and Typing via Linear Logic

R\'emi Di Guardia , Thomas Ehrhard , Claudia Faggian

Authors on Pith no claims yet

Pith reviewed 2026-05-07 12:20 UTC · model grok-4.3

classification 💻 cs.LO

keywords bayesiannetworksquantumclassicalcompositionalformalismlinearlogic

0 comments

The pith

Quantum Bayesian networks receive a compositional semantics and linear-logic typing that reduces to standard Bayesian networks for classical data and to tensor networks for quantum data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Bayesian networks are diagrams that show how causes lead to effects and let us calculate probabilities. The classical version uses factors and multiplication to combine information. Quantum versions must also handle superposition and entanglement. This work creates a single framework that works for both. When every variable is ordinary classical data, the new rules give exactly the same numbers as Pearl's original networks. When everything is quantum, the rules become the same as tensor networks used in quantum computing. To control how pieces are put together, the authors use proof-nets from linear logic. These nets carry types that act like labels saying which parts can be connected without breaking the rules. The result is a typed language for building and reasoning about systems that mix classical probabilities with quantum measurements.

Core claim

A key feature of our compositional semantics is that when all causes are classical, it coincides with the standard factor-based semantics of Bayesian networks, while in the purely quantum case it reduces to tensor networks. We then propose a typed formalism based on linear logic proof-nets, where types ensure well-behaved composition of systems.

Load-bearing premise

That the proposed linear-logic typing and compositional semantics correctly capture all causal relations and measurement probabilities in mixed classical-quantum systems without requiring extra constraints or losing expressiveness.

read the original abstract

Quantum Bayesian networks provide a mathematical formalism to describe causal relations, to analyse correlations, and to predict the probabilities of measurement outcomes, in systems involving both classical and quantum data. They generalize Pearl's Bayesian networks-prominent graphical models for classical probabilistic reasoning and inference. Our paper brings compositional principles and a typing discipline into this setting. A key feature of our compositional semantics is that when all causes are classical, it coincides with the standard factor-based semantics of Bayesian networks, while in the purely quantum case it reduces to tensor networks. We then propose a typed formalism based on linear logic proof-nets, where types ensure well-behaved composition of systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper sketches a compositional semantics for quantum Bayesian networks that reduces to classical factor graphs or tensor networks by construction, then adds linear-logic proof-net typing, but the abstract shows none of the actual math or examples.

read the letter

The main thing to know is that the authors define a compositional semantics for quantum Bayesian networks whose key property is that it matches ordinary Bayesian network factor semantics when all variables are classical and reduces to tensor networks when everything is quantum. They then propose to control composition with a linear-logic typing discipline based on proof-nets. That interpolation plus the typing step is the new piece; it is not described in the prior literature they cite on either classical networks or pure tensor networks. The construction is presented as recovering the known cases directly rather than by fitting, which is a reasonable design choice if the details work out. The abstract states the reductions and the typing goal clearly, which is helpful even at this level. The main limitation is that nothing beyond the high-level claim is shown. There are no proofs, no small worked examples, no derivation of the semantics, and no check that mixed classical-quantum systems are handled without extra constraints or loss of expressiveness. That leaves the central claim unverified from what we have. The citation pattern is standard and appropriate for the area, but the paper does not yet engage with concrete prior results on quantum causal models in enough depth to judge overlap. This is aimed at researchers who already work with both Bayesian networks and categorical or linear-logic approaches to quantum information. Someone looking for a typed framework that unifies the two settings could find it useful once the details are filled in. It deserves a serious referee. The idea is novel enough and the potential application clear enough that referees should see the full development rather than a desk rejection.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract relies on standard background from linear logic and quantum mechanics but introduces no explicit free parameters, new physical entities, or ad-hoc axioms beyond the claim that the new semantics coincides with known models in the classical and quantum limits.

pith-pipeline@v0.9.0 · 5403 in / 1086 out tokens · 54903 ms · 2026-05-07T12:20:20.711944+00:00 · methodology

Quantum Bayesian Networks: Compositionality and Typing via Linear Logic

Core claim

Load-bearing premise

discussion (0)