Reproduced Papers

MerLin provides reproducible implementations of published quantum machine learning papers. Each card links to a dedicated reproduction page with paper metadata, implementation details, code access, and results.

Available Reproductions

The reproductions are organized by topic. Each card opens the corresponding paper-reproduction page.

Kernel Methods

Photonic Quantum-Enhanced Kernels

Kernel-based classification on a photonic processor with indistinguishable-versus-distinguishable photon comparisons.

Yin et al.2025CompleteKernel methods

Nearest Centroid Quantum Classification

Amplitude-encoding nearest-centroid classification with quantum inner-product estimation.

Johri et al.2020CompleteImage TBD

For a Better Understanding of Photonic QML Theory

Fock State-Enhanced Expressivity

Expressivity study of photonic variational models with increasing photon counts.

Gan et al.2022Complete

HQNN: Myth or Reality?

Controlled benchmark of hybrid quantum neural networks versus classical baselines across feature complexity.

Kashif et al.2025Complete

Experimental Data Re-uploading

Resource-efficient photonic data re-uploading model reproduced on multiple datasets.

Mauser et al.2025Complete

Limitations of Amplitude Encoding

Theoretical and experimental study of amplitude-encoding concentration effects and their impact on quantum classification.

Wang et al.2025Complete

Computer Vision

Quantum Optical Reservoir Computing

Boson-sampling reservoirs for image recognition and scalable photonic reservoir analysis.

Sakurai et al.2025In Progress

Photonic QCNN with State Injection

Photonic quantum convolutional architecture with adaptive state injection for vision tasks.

Monbroussou et al.2025Complete

QCNN for Classical Data Classification

Quantum pseudo-convolution workflow for PCA-compressed classical image classification.

Hur et al.2022Complete

Sequential Tasks

Quantum Recurrent Neural Networks for Sequential Learning

QRNN for NISQ devices with photonic implementation and evaluation on sequential tasks.

Li et al.2023CompletePennylaneSequential

Quantum Long Short-Term Memory

Quantum LSTM architecture with photonic implementation for sequence and temporal modeling tasks.

Yen-Chi et al.2020CompleteSequential

Advanced Training Paradigms

Quantum LLM Fine-Tuning

Photonic quantum heads for sentence-embedding classification with baseline-aware evaluation.

Kim et al.2025Complete

Quantum Self-Supervised Learning

SimCLR-style qSSL reproduction with MerLin photonic, Qiskit, and classical backends.

Jaderberg et al.2022Complete

Quantum Adversarial Machine Learning

Adversarial attacks and defenses for hybrid quantum-classical photonic classifiers.

Lu et al.2020CompleteImage TBD

Photonic Quantum GAN

Photonic QGAN for classical image generation with a MerLin QuantumLayer patch generator.

Sedrakyan and Salavrakos2025In Progress

Quantum Transfer Learning

Hybrid transfer learning with frozen classical feature extractors and trainable quantum heads.

Mari et al.2020CompleteTransfer learning

Quantum Relational Knowledge Distillation

Knowledge distillation (KD) is a widely adopted technique for compressing large models into smaller, more efficient student models

Liu et al.2025CompleteKnowledge Distillation

Distributed Training

Distributed Quantum Neural Networks

Distributed photonic parameter generation for CNN compression and MNIST-classification experiments.

Chen et al.2025Complete

Future-proofing

Quantum Memristor Neuromorphic Computing

Photonic quantum-memristor reservoir model with reproduced nonlinear-task improvements.

Selimovic et al.2025Partial

Contributing Reproductions

We welcome contributions of additional paper reproductions.

Requirements:

High-impact quantum ML papers (>50 citations preferred)
Photonic/optical quantum computing focus
Implementable with current MerLin features
Clear experimental validation

Submission Process:

Propose the paper in our GitHub Discussions
Implement using MerLin following our guidelines
Validate results against original paper
Document in Jupyter notebook format
Submit via pull request a complete reproduction folder and a summary page in docs/source/reproduced_papers/reproductions/ directory

Mandatory Structure for a Reproduction:

papers/NAME/            # Non-ambiguous acronym or fullname of the reproduced paper
├── .gitignore          # specific .gitignore rules for clean repository
├── notebook.ipynb      # Interactive exploration of key concepts
├── README.md           # Paper overview and results overview
├── requirements.txt    # additional requirements for the scripts
├── configs/            # defaults + CLI/runtime descriptors consumed by the repo root runner
├── lib/                # code used by the shared runner and notebooks - as an integrated library (import shared data helpers from papers/shared/<paper>/)
├── models/             # Trained models
├── results/            # Selected generated figures, tables, or outputs from trained models
├── tests/              # Validation tests
└── utils/              # additional commandline utilities for visualization, launch of multiple trainings, etc...

Template Summary Page: this document

Recognition

Contributors to reproductions are recognized in:

Paper reproduction documentation
MerLin project contributors list
Academic citations in MerLin publications

Have a paper you’d like to see reproduced? Start a discussion.