🔬 PennyLane Demonstrations — Agent Experiments Fork

Agent-driven experiments fork of the PennyLane demonstration suite.
Contains 5 Python implementations of testable hypotheses in hybrid quantum-classical ML.
_{⬆️ Inherits the full upstream demo collection from PennyLaneAI/demos.}

_{Fork architecture showing upstream origin, added experiments, and mapping of each hypothesis to its PennyLane baseline demo. Each experiment extends a specific tutorial.}

---

🌐 Ecosystem Overview

graph TB
    classDef north fill:#1a1a2e,stroke:#00d4aa,stroke-width:2,color:#ccc
    classDef impl fill:#1a1a2e,stroke:#7C3AED,stroke-width:3,color:#fff
    classDef exec fill:#1a1a2e,stroke:#10B981,stroke-width:2,color:#ccc
    classDef up fill:#333,stroke:#666,stroke-width:1,color:#999

    subgraph Upstream_Source["☁️ Upstream"]
        U[("🔬 PennyLaneAI/demos<br/>Official Demos")]
    end
    subgraph Agent_Stack["🤖 Agent Experiment Stack"]
        direction TB
        H[("🧪 agent-hypotheses<br/>Research Definitions")]
        D[("🐍 pennylane-demos<br/><b>← You Are Here</b>")]
        E[("📓 agent-experiments<br/>Runnable Notebooks")]
    end
    subgraph Deliverables["📦 Deliverables"]
        S[("📁 experiments/<br/>H1–H5 Python scripts")]
        R[("📊 results/<br/>Metrics & Plots")]
    end

    U -.->|"fork + extend"| D
    H -->|"defines"| D
    D -->|"implements"| S
    S -->|"packaged as"| E
    S --> R
    E -->|"results validate"| H

    class H north
    class D,impl,S,R impl
    class E exec
    class U up

Loading

---

🧪 What's New: Agent Experiments (experiments/)

This fork adds 5 agent-generated experiment scripts on top of the full PennyLane demo suite.

Experiment Matrix

graph LR
    subgraph H1_H2["Trainability Track"]
        H1["H1: NAS HQNN"]
        H2["H2: Dissipation vs Local Cost"]
    end
    subgraph H3_H5["Expressivity Track"]
        H3["H3: Post-Variational"]
        H4["H4: PDE-Constrained"]
        H5["H5: Data-Reuploading"]
    end
    H1_H2 -->|"BP mitigation"| H3_H5
    style H1 fill:#8B5CF6,color:#fff
    style H2 fill:#3B82F6,color:#fff
    style H3 fill:#EF4444,color:#fff
    style H4 fill:#F59E0B,color:#fff
    style H5 fill:#10B981,color:#fff

Loading

ID	Script	Area	Est. Cost	Dependencies	Run Command
H1	h1_joint_nas_hqnn.py			pennylane torch scikit-learn	python experiments/h1_joint_nas_hqnn.py
H2	h2_engineered_dissipation_bp.py			pennylane matplotlib	python experiments/h2_engineered_dissipation_bp.py
H3	h3_post_variational_benchmarks.py			pennylane scikit-learn	python experiments/h3_post_variational_benchmarks.py
H4	h4_pde_constrained_loss.py			pennylane matplotlib	python experiments/h4_pde_constrained_loss.py
H5	h5_data_reuploading_scaling.py			pennylane scikit-learn	python experiments/h5_data_reuploading_scaling.py

Each script is self-contained with CLI argument parsing (--help for options) and timestamped JSON output saved to experiments/results/.

---

🚀 Getting Started

Prerequisites

# Python 3.10+
python --version

# PennyLane (core + lightning simulator)
pip install pennylane pennylane-lightning

Run All Experiments

# Clone this fork
git clone https://github.com/NullLabTests/pennylane-demos.git
cd pennylane-demos

# Run experiments sequentially
for script in experiments/h*.py; do
    echo "=== Running $script ==="
    python "$script"
    echo ""
done

Run with Custom Parameters

Each script supports CLI arguments:

python experiments/h1_joint_nas_hqnn.py --n-samples 200 --n-epochs 20 --output-dir experiments/results
python experiments/h2_engineered_dissipation_bp.py           # uses defaults
python experiments/h3_post_variational_benchmarks.py        # uses defaults
python experiments/h4_pde_constrained_loss.py               # uses defaults
python experiments/h5_data_reuploading_scaling.py           # uses defaults

Results are saved to experiments/results/<id>_<timestamp>.json with plots in PNG format.

---

📁 Repository Structure

📦 pennylane-demos (fork)
│
├── 🧪 AGENT EXPERIMENTS — Added content
│   ├── 📁 experiments/
│   │   ├── h1_joint_nas_hqnn.py              # H1: NAS for Pareto-optimal HQNNs
│   │   ├── h2_engineered_dissipation_bp.py    # H2: Engineered dissipation vs local cost
│   │   ├── h3_post_variational_benchmarks.py  # H3: Post-variational strategies
│   │   ├── h4_pde_constrained_loss.py         # H4: PDE-constrained loss functions
│   │   ├── h5_data_reuploading_scaling.py     # H5: Trainable data reuploading
│   │   └── 📁 results/                        # 📊 Generated metrics & plots (gitignored)
│   └── 📄 README.md                           # ℹ️ This file
│
├── ⬆️ UPSTREAM CONTENT — Inherited from PennyLaneAI/demos
│   ├── 📁 demonstrations_v2/                  # 🌐 Official PennyLane demo notebooks
│   ├── 📁 dependencies/                       # 📦 Dependency metadata
│   ├── 📁 documentation/                      # 📖 CLI tool & docs
│   ├── 📁 lib/                                # 🛠️ Shared utilities
│   ├── 📁 _static/                            # 🎨 Web assets
│   ├── 📁 _templates/                         # 📝 Sphinx templates
│   ├── 📄 conf.py                             # ⚙️ Sphinx configuration
│   ├── 📄 index.rst                           # 📑 Documentation index
│   ├── 📄 CONTRIBUTING.md                     # 📖 Contributor guide
│   └── 📄 LICENSE                             # ⚖️ Apache 2.0
│
└── 🛠️ BUILD & CI
    ├── 📄 pyproject.toml                      # 🐍 Poetry project config
    ├── 📄 poetry.lock                         # 🔒 Locked dependencies
    ├── 📄 extension.py                        # 🔧 Sphinx extension
    └── 📁 .github/                            # 🤖 CI workflows

---

🔗 Sister Repositories

Repository	Badge	Purpose	Link
agent-hypotheses		Research hypotheses & metrics	→ View
pennylane-demos (this repo)		Python scripts (forked)	← You are here
agent-experiments		Notebooks + CI	→ View

Data Flow

flowchart LR
    subgraph Up[" "]
        U["PennyLaneAI/demos"] -.->|"basis for"| D["pennylane-demos"]
    end
    subgraph Core[" "]
        H["agent-hypotheses"] -->|"informs"| D
    end
    subgraph Down[" "]
        D -->|"wraps into"| E["agent-experiments"]
    end
    style U fill:#333,stroke:#666,color:#999,stroke-width:1
    style H fill:#1a1a2e,stroke:#00d4aa,color:#fff,stroke-width:2
    style D fill:#1a1a2e,stroke:#7C3AED,color:#fff,stroke-width:3
    style E fill:#1a1a2e,stroke:#10B981,color:#fff,stroke-width:2

Loading

_{End-to-end data flow: hypotheses.json defines experiment parameters → Python scripts execute with CLI args → timestamped JSON results validate hypotheses. A complete feedback loop across all three repositories.}

---

⬆️ Upstream Attribution

This fork inherits the full PennyLane demonstration suite — a collection of tutorials and implementations ranging from introductory concepts to cutting-edge quantum computing research. Built by Xanadu, for research.

Resource	Link
🌐 Online Demonstrations	pennylane.ai/qml/demonstrations
📖 Contributing Guide	CONTRIBUTING.md
🛠️ CLI Tool	documentation/demo-cli.md
📦 Dependency Management	dependencies/README.md
🐛 Issue Tracker	PennyLaneAI/demos/issues

---

📄 License

The materials and demonstrations contained within this repository are free and open-source, released under the Apache License, Version 2.0.

Please note, the file custom_directives.py is available under the BSD 3-Clause License, with copyright © 2017, PyTorch contributors.

---

_{Part of the NullLabTests agent-driven research ecosystem.}