In this paper, we study the online quantum state learning problem which is recently proposed by Aaronson et al. (2018). In this problem, the learning algorithm sequentially predicts quantum states based on observed measurements and losses and the goal is to minimize the regret. In the previous work, the existing algorithms may output mixed quantum states. However, in many scenarios, the prediction of a pure quantum state is required. In this paper, we first propose a Follow-the-Perturbed-Leader (FTPL) algorithm that can guarantee to predict pure quantum states. Theoretical analysis shows that our algorithm can achieve an O(√T) expected regret under some reasonable settings. In the case that the pure state prediction is not mandatory, we propose another deterministic learning algorithm which is simpler and more efficient. The algorithm is based on the online gradient descent (OGD) method and can also achieve an O(√T) regret bound. The main technical contribution of this result is an algorithm of projecting an arbitrary Hermitian matrix onto the set of density matrices with respect to the Frobenius norm. We think this subroutine is of independent interest and can be widely used in many other problems in the quantum computing area. In addition to the theoretical analysis, we evaluate the algorithms with a series of simulation experiments. The experimental results show that our FTPL method and OGD method outperform the existing RFTL approach proposed by Aaronson et al. (2018) in almost all settings. In the implementation of the RFTL approach, we give a closed-form solution to the algorithm. This provides an efficient, accurate, and completely executable solution to the RFTL method.

Due to the decoherence of the state-of-the-art physical implementations of quantum computers, it is essential to parallelize the quantum circuits to reduce their depth. Two decades ago, Moore and Nilsson [1] demonstrated that additional qubits (or ancillae) could be used to design “shallow” parallel circuits for quantum operators. They proved that any n-qubit CNOT circuit could be parallelized to O(log n) depth, with O(n2) ancillae. However, the near-term quantum technologies can only support limited amount of qubits, making space-depth trade-off a fundamental research subject for quantum-circuit synthesis. In this work, we establish an asymptotically optimal space-depth trade-off for the design of CNOT circuits. We prove that for any m ≥ 0, any n-qubit CNOT circuit can be parallelized to depth, with m ancillae. We show that this bound is tight by a counting argument, and further show that even with arbitrary two-qubit quantum gates to approximate CNOT circuits, the depth lower bound still meets our construction, illustrating the robustness of our result. Our work improves upon two previous results, one by Moore and Nilsson [1] for O(log n)-depth quantum synthesis, and one by Patel, Markov, and Hayes [2] for m =0: for the former, we reduce the need for ancillae by a factor of log2 n by showing that m = O(n2 / log2 n) additional qubits — which is asymptotically optimal — suffice to build O(log n)-depth, O(n2 / log n)-size CNOT circuits; for the later, we reduce the depth by a factor of n to the asymptotically optimal bound . Our results can be directly extended to stabilizer circuits using an earlier result by Aaronson and Gottesman [3]. In addition, we provide relevant hardness evidence for synthesis optimization of CNOT circuits in term of both size and depth.

Reversible Boolean function (RBF) is a one-to-one function which maps n-bit input to n-bit output. Reversible logic synthesis has been widely studied due to its connection with low-energy computation as well as quantum computation. In this paper, we give a structured decomposition for even RBFs. Specifically, for n ≥ 6, any even n-bit RBF can be decomposed to 7 blocks of (n-1)-bit RBF, where 7 is a constant independent of n and the positions of these blocks have a large degree of freedom. Moreover, if the (n-1)-bit RBFs are required to be even as well, we show for n ≥ 10, even n-bit RBF can be decomposed to 10 even (n - 1)-bit RBFs. In short, our decomposition has block depth 7 and even block depth 10. Our result improves Selinger's work in block depth model, by reducing the constant from 9 to 7 and from 13 to 10, when the blocks are limited to be even. We emphasize that our setting is a bit different from Selinger's work. In Selinger's constructive proof, each block is placed in one of two specific positions and thus the decomposition has an alternating structure. We relax this restriction and allow each block to act on arbitrary (n - 1) bits. This relaxation keeps the block structure and provides more candidates when choosing the positions of blocks.