This work advances the theoretical understanding of quantum learning by establishing a new family of upper bounds on the expected generalization error of quantum learning algorithms, leveraging the framework introduced by Caro et al. (2024) and a new definition for the expected true loss. Our primary contribution is the derivation of these bounds in terms of quantum and classical Rényi divergences, utilizing a variational approach for evaluating quantum Rényi divergences, specifically the Petz and a newly introduced modified sandwich quantum Rényi divergence. Analytically and numerically, we demonstrate the superior performance of the bounds derived using the modified sandwich quantum Rényi divergence compared to those based on the Petz divergence. Furthermore, we provide probabilistic generalization error bounds using two distinct techniques: one based on the modified sandwich quantum Rényi divergence and classical Rényi divergence, and another employing smooth max Rényi divergence.

We study two kinds of different problems. One is the multiple independence testing, which can be considered as a kind of generalization of quantum Stein’s lemma. We test whether the quantum system is correlated to the classical system or is independent of it. Here, the null hypothesis is composed of states having the quantum system is correlated to the classical system in an arbitrarily varying form. The second problem is the problem of reliable communication over classical-quantum arbitrarily varying multiple access channels (CQ-AVMAC) and establishing its capacity region by giving multiple achievability techniques. We prove that each of these techniques is optimal by proving a converse. Further, for both these techniques, the decoder designed is a universal decoder and can achieve any rate pair in the capacity region without time sharing and also these decoders do not depend on the channel and therefore they are universal. Our result covers the case when the channel parameter is continuous, which has not been studied even in the classical case. Further, both these techniques can be easily generalized to the case when there are$T (T\gt 2)$senders. The design of each of these decoders is based on the study of multiple independence testing. This approach allows us to study the problem of reliable communication over CQ-AVMAC from the point of view of hypothesis testing. Further, we also give a necessary and sufficient condition for the deterministic code capacity region of CQ-AVMAC to be non-empty.

In information theory, we often use intersection and union of the typical sets to analyze various communication problems. However, in the quantum setting it is not very clear how to construct a measurement which behaves analogously to intersection and union of the typical sets. In this work, we construct a projection operator which behaves very similarly to intersection and union of the typical sets. Our construction relies on the Jordan’s lemma. Using this construction we study the problem of communication over authenticated classical-quantum channels and derive its capacity. As another application of our construction, we also study the problem of quantum asymmetric composite hypothesis testing.