cv-photonic-qnn-edge-medical

name	cv-photonic-qnn-edge-medical
description	Parameter-efficient continuous-variable photonic quantum neural networks for edge medical AI. Simplified Phi-D-U1 CV-QNN architecture cuts trainable parameters 40-45%, mitigates barren plateaus, achieves 100% calibrated test accuracy with 18 parameters for oral cancer detection. Use when: CV quantum neural network design, photonic quantum ML, medical image classification on edge devices, barren plateau mitigation, parameter-efficient quantum classifiers, room-temperature quantum computing.
metadata	{"arxiv_id":"2606.28252","published":"2026-06-26","authors":"Akshay Bhagwan Sonawane, Sophie Choe, Lakshman Tamil","tags":["CV-QNN","photonic-quantum","medical-AI","edge-computing","oral-cancer","barren-plateau"]}

CV Photonic QNN for Edge Medical AI

Core Concept

Hybrid classical-continuous-variable quantum neural network for parameter-efficient medical image classification at the edge. Uses room-temperature photonic quantum computing (vs. cryogenic qubit hardware) combined with MobileNetV1 feature extraction and PCA dimensionality reduction to achieve medical-grade accuracy with minimal trainable quantum parameters.

Architecture

Pipeline

1. Feature extraction: MobileNetV1 pretrained classical backbone → smartphone image features
2. Dimensionality reduction: PCA to 16 dimensions
3. Quantum encoding: Angle encoding into CV quantum states
4. CV-QNN layers: Displacement (D), interferometric (U1), and Kerr nonlinear gates on photonic backend

Simplified Layer Design

• Standard CV-QNN (Killoran et al. 2019): Full D → S → R → U → K gate sequence
• Simplified Phi-D-U1: Reduced Phi ∘ D ∘ U1 sequence cutting trainable parameters by 40-45%