quantum-neural-entropy-estimation

name	quantum-neural-entropy-estimation
description	Entropy estimation in multi-qutrit quantum systems using variational quantum algorithms and classical CNNs. Combines SU(3)-inspired ansatzes for VQAs on small systems with CNN-based estimators for larger qutrit systems. Use when: (1) estimating von Neumann entropy of quantum states, (2) designing quantum state tomography alternatives, (3) building hybrid quantum-classical ML pipelines for quantum characterization, (4) comparing VQA vs classical neural network approaches for quantum system analysis. Triggers: quantum entropy, qutrit, SU(3) ansatz, state tomography, variational quantum algorithm, CNN quantum, mutual unbiased bases, quantum state estimation, quantum machine learning characterization.
metadata	{"arxiv_id":"2606.20504","published":"2026-06-18","authors":"Sai Sakunthala Guddanti, Anil Prabhakar, Ria Rushin Joseph","tags":["quantum","neural-networks","entropy-estimation","variational-algorithms","cnn","qutrit","quantum-ml"]}

Quantum Neural Entropy Estimation

Methodology for von Neumann entropy estimation in multi-qutrit systems using VQAs (small systems) and CNNs (larger systems). arXiv:2606.20504

Core Methodology

Two-Regime Approach

• Small systems (2-3 qutrits): Use variational quantum algorithms with SU(3)-inspired hardware-efficient ansatzes
• Large systems (4-5+ qutrits): Use classical CNNs trained on measurement outcomes from tensor-product mutually unbiased bases (MUBs)

Key Findings

1. VQA accuracy depends primarily on trainable parameter count (~120 is optimal), not entangling gate count beyond a threshold
2. CNN estimators achieve 90th-percentile absolute errors of ~0.13-0.16 nats using only 12.5% of full tomography measurements
3. CNN models are robust to shot noise and generalize to out-of-distribution states