Title:Conditional entropy and information of quantum processes

Abstract:What would be a reasonable definition of the conditional entropy of bipartite quantum processes, and what novel insight would it provide? We develop this notion using four information-theoretic axioms and define the corresponding quantitative formulas. Our definitions of the conditional entropies of channels are based on the generalized state and channel divergences, such as quantum relative entropy, min- and max-relative entropy, etc. We show that the conditional entropy of quantum channels can potentially reveal important features of the channel, such as its underlying causal structure, which cannot be captured by the entropy of quantum channels or the conditional entropy of bipartite states. Specifically, if the von Neumann conditional entropy $S[A|B]_{\mathcal{N}}$ of a quantum channel $\mathcal{N}_{A'B'\to AB}$ is strictly less than $-\log|A|$, then the channel necessarily has causal influence from $A'$ to $B$. Furthermore, we show that our definition of conditional entropy establishes the strong subadditivity of the entropy for quantum channels. We study the total amount of correlations possible due to quantum processes by defining the multipartite mutual information of quantum channels.