arXiv:2610.29373
Convergence Analysis of the Quantum Blahut-Arimoto Algorithm and the Contraction Coefficient of Qubit Amplitude-Damping Channels
Junke Han, Yonglong Li
Abstract
In this work, we investigate the convergence rate of the quantum Blahut-Arimoto (BA) algorithm proposed for computing the coherent information of less noisy quantum channels. Previous work established exponential convergence under certain restrictive assumptions on the channel. Using a Taylor expansion around a positive definite optimizer, we show that, for an injective degradable channel whose degrading channel has a relative-entropy contraction coefficient strictly less than one, the algorithm converges at an exponential rate after sufficiently many iterations. We further verify this condition for qubit amplitude-damping (AD) channels in the relevant parameter regime. In addition, by analyzing the Bogoliubov-Kubo-Mori (BKM) metric on the Bloch ball, we derive an explicit characterization of the relative-entropy contraction coefficient of qubit AD channels. The proposed approach is further extended to calculate the relative contraction coefficient and the relative expansion coefficient of two qubit AD channels.
AI Summary
The paper studies when the quantum Blahut–Arimoto (BA) algorithm for maximizing coherent information converges exponentially, despite the failure of previously used global relative-entropy expansion assumptions. It also computes relative-entropy contraction and channel-comparison coefficients for qubit amplitude-damping (AD) channels. For an injective degradable channel $\Phi$ with degrading map $\mathcal D$ satisfying $\eta(\mathcal D)<1$, coherent information is strictly concave and has a unique optimizer. If that optimizer $\rho^\star$ is full rank, BA converges exponentially after sufficiently many iterations: its objective gap is bounded by $\gamma(1-a/\gamma)^{t-N_0}D(\rho^\star\|\rho^{(N_0)})$. The proof supplies $a=\lambda_{\min}(\rho^\star)\lambda_{\min}(\Phi^\dagger\Phi)[1-\eta(\mathcal D)]/4$, combining convergence to the optimizer, a local quadratic bound via quantum $\chi^2$ divergence, injectivity, and strong data processing. The assumptions are verified for qubit dephasing away from complete dephasing and AD with damping $0\le p<1/2$. For AD, Bloch-ball symmetry and comparison of BKM-metric quadratic forms reduce the contraction coefficient exactly to $\eta(\mathcal A_p)=\sup_{0\le r<1}(1-p)\frac{r}{\operatorname{arctanh}r}\frac{\operatorname{arctanh}[p+(1-p)r]}{p+(1-p)r}$. An analogous restricted one-variable formula determines relative contraction between two AD channels. Their relative expansion coefficient has the closed form $\check\eta(\mathcal A_{p_2},\mathcal A_{p_1})=p_1(1-p_2)/[p_2(1-p_1)]$ for $0<p_1<p_2<1$. The convergence guarantee is asymptotic rather than a global exponential bound, requires a full-rank optimizer and the stated channel conditions, and does not quantify the entry time $N_0$. The contraction characterization remains a scalar optimization rather than a closed-form value.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 44 pages