{"id":"arxiv/2607.27051","url":"https://arxiv.org/abs/2607.27051","tags":["quant-ph"],"score":0,"title":"Quasi-polar Decomposition of Quantum Neural Networks via Adaptive Non-local Observables","domain":"quant-ph","source":"arxiv","authors":["Shih-Hao Ho","Yan Li","Huan-Hsin Tseng","Hsin-Yi Lin","Samuel Yen-Chi Chen","Shinjae Yoo"],"reviews":0,"subject":"quant-ph","summary":"We use Diagonal Adaptive Non-local Observables (DANO) as a canonical decomposition for studying Variational Quantum Circuit model evolution. Separating each learned observable into a diagonal spectrum and a unitary basis gives a quasi-polar description: the spectral weights are viewed as radial coordinates, while the unitary circuit serves as angular coordinates through Lie group identifications. This turns the training process into a trajectory in spectral and Lie-algebra space. Experiments on two classification tasks show that DANO radial spectral expansion correlates with accuracy. DANO…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-07-29","reactions":{}}