Liqi Wang

Zhiwei Wang, Liqi Wang

The minimum-weight perfect matching (MWPM) decoder is a standard decoding strategy for surface codes, but its performance degrades considerably under biased noise. In this paper, a modified surface code, termed the XYZ planar code, is introduced, and the MWPM decoder is extended to posterior MWPM (pMWPM) with almost no increase in decoding complexity. The XYZ planar code exhibits higher and more stable thresholds than the planar code under almost all bias conditions, while also achieving significantly lower logical error rates. Specifically, in the infinite-bias case, the threshold of the XYZ planar code is improved by about \(36\%\) compared to that of the surface code, and it maintains comparable or higher thresholds under other biases -- for example, the threshold reaches approximately \(15.5\%\) at bias \(η= 1\) and \(14.2\%\) at \(η= 100\). Furthermore, pMWPM can be adapted to a wide range of modified surface codes, and the results presented in this work also indicate its excellent potential in other scenarios, such as configurations in which \(Y\) operators involve a larger number of data qubits.

Liqi Wang, Qiya Hu

Dropout as a regularization technique is widely used in fully connected layers while is less effective in convolutional layers. Therefore more structured forms of dropout have been proposed to regularize convolutional networks. The disadvantage of these methods is that the randomness introduced causes inconsistency between training and inference. In this paper, we apply a mutual learning training strategy for convolutional layer regularization, namely R-Block, which forces two outputs of the generated difference maximizing sub models to be consistent with each other. Concretely, R-Block minimizes the losses between the output distributions of two sub models with different drop regions for each sample in the training dataset. We design two approaches to construct such sub models. Our experiments demonstrate that R-Block achieves better performance than other existing structured dropout variants. We also demonstrate that our approaches to construct sub models outperforms others.