quantum-predicate-learning-sgg

name	quantum-predicate-learning-sgg
description	QPredSGG: Hybrid Quantum Predicate Learning for Long-Tailed Scene Graph Generation. Use when analyzing quantum algorithms, complexity bounds, quantum ML architectures, or quantum error correction involving mathematical analysis and statistical methods.
metadata	{"arxiv_id":"2606.04689","published":"2026-06-06","category":"quantum-ml"}

QPredSGG: Hybrid Quantum Predicate Learning for Long-Tailed Scene Graph Generation

Core Methodology

Scene Graph Generation (SGG) requires relational reasoning over objects and their interactions, but performance is limited by severe long-tail predicate imbalance. This work introduces a hybrid quantum predicate classifier for SGG by replacing the classical predicate head in Causal Feature Enhancement Network (CFEN) with a Quantum Predicate Head (QP-Head). The best 4-qubit QP-Head uses Amplitude Embedding and Strongly Entangling Layers to compress 4096-dimensional pair features into 16-dimensional quantum-compatible representation (256x reduction), achieving mR@100 of 57.25% vs 41.1% classical CFEN reference with only 96 trainable quantum parameters. Demonstrates parameter-efficient long-tail relational classification in visual reasoning.

Key Mathematical Framework

• Domain: quantum-ml
• arXiv: 2606.04689
• Date: 2026-06-06
• Math Keywords: amplitude embedding, entangling layers, dimensionality reduction, cross-entropy optimization

Application Patterns

Pattern 1: Mathematical Analysis

• Identify core mathematical structures in quantum protocols
• Map to complexity theory bounds or statistical models
• Extract reusable analytical patterns

Pattern 2: Quantum-Classical Comparison

• Compare quantum vs classical performance metrics
• Quantify parameter efficiency gains
• Analyze scaling behavior

Activation Keywords

• 2606.04689
• quantum predicate learning, scene graph generation, long-tail classification, amplitude embedding, quantum neural network, hybrid quantum-classical