Tag
该论文证明多项式图滤波器（线性消息传递与谱图网络的核心）在长程伊辛模型基准上存在无法通过提高阶数消除的误差下界，并提出分层步长滤波器与膨胀卷积网络，显著超越已发表的最佳结果。
该论文提出ΣΔ-Sampler优化器，将二值和三值神经网络训练损失映射为高阶Ising模型能量，并通过噪声整形与退火实现高效的边缘设备训练，在CIFAR-10、ImageNet-1k和MatMul-free语言模型上验证了接近全精度的精度，同时将训练内存减少超过50%、速度提升一个数量级。
IBM announces three new entries in its quantum advantage tracker, each using different approaches to overcome errors and validate quantum results, demonstrating quantum advantage in ways that can be trusted even when classical verification is infeasible.
Proposes Trans-Ising, a transfer learning method for high-dimensional Ising models that uses a loss-based source screening rule and two-stage estimation to improve estimation accuracy over target-only and naive pooling methods.
This paper presents a scalable backpropagation-based algorithm for training deep convolutional networks to run on thermodynamic Ising hardware, achieving 94.9% on CIFAR-10 and 76.0% on CIFAR-100 while analyzing inference cost-accuracy tradeoffs.