Chenyu Qiao | Searcharxiv

arXiv2025

High-Efficiency Quantum Memory of Full-Bandwidth Squeezed Light

In continuous-variable quantum information processing, it is crucial to develop high-efficiency and broadband quantum memory of squeezed light, which enables the storage of full-bandwidth information. Here, we present a quantum memory of squeezed light with up to 24 MHz bandwidth, which is at least 12 times that of previous narrowband resonant memory systems, via a far-off resonant Raman process. We achieve output squeezing of as high as 1.0 dB with fidelity above 92% and a memory efficiency of 80%, corresponding to an end-to-end efficiency of 64.2%, when input squeezing is 1.6 dB. The lowest excess noise of 0.025 shot-noise-unit in the memory system is estimated by the noisy channel model which is benefited from optimizing quantum memory performance with a backward retrieval strategy. Our results represent a breakthrough in high-performance memory for squeezed states within tens of MHz-level bandwidth, which has potential applications in high-speed quantum information processing.

quant-ph↗

arXiv2024

Adaptive Multi-Modal Cross-Entropy Loss for Stereo Matching

Despite the great success of deep learning in stereo matching, recovering accurate disparity maps is still challenging. Currently, L1 and cross-entropy are the two most widely used losses for stereo network training. Compared with the former, the latter usually performs better thanks to its probability modeling and direct supervision to the cost volume. However, how to accurately model the stereo ground-truth for cross-entropy loss remains largely under-explored. Existing works simply assume that the ground-truth distributions are uni-modal, which ignores the fact that most of the edge pixels can be multi-modal. In this paper, a novel adaptive multi-modal cross-entropy loss (ADL) is proposed to guide the networks to learn different distribution patterns for each pixel. Moreover, we optimize the disparity estimator to further alleviate the bleeding or misalignment artifacts in inference. Extensive experimental results show that our method is generic and can help classic stereo networks regain state-of-the-art performance. In particular, GANet with our method ranks $1^{st}$ on both the KITTI 2015 and 2012 benchmarks among the published methods. Meanwhile, excellent synthetic-to-realistic generalization performance can be achieved by simply replacing the traditional loss with ours.

cs.CV↗