Metadata-Version: 2.4
Name: quantlof
Version: 0.1.5
Summary: Quantum-enhanced Local Outlier Factor algorithm
Author: Yuto Mueller
Author-email: geoyuto@gmail.com
License: MIT
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE.m
Requires-Dist: qiskit==2.0.2
Requires-Dist: qiskit-aer==0.17.0
Requires-Dist: qiskit-ibm-runtime==0.40.0
Requires-Dist: scikit-learn==1.6.1
Requires-Dist: scipy==1.15.3
Requires-Dist: numpy==2.2.6
Dynamic: author
Dynamic: author-email
Dynamic: description
Dynamic: description-content-type
Dynamic: license
Dynamic: license-file
Dynamic: requires-dist
Dynamic: requires-python
Dynamic: summary

# QuantumLOFClassifier – Inspired by Guo et al. (2023)

This repository provides an implementation partially inspired by the algorithm proposed in:

> Ming-Chao Guo et al., “Quantum Algorithm for Unsupervised Anomaly Detection”
> [arXiv:2304.08710 (2023)](https://arxiv.org/abs/2304.08710)

## 📘 Overview

`QuantumLOFClassifier` is a quantum-enhanced Local Outlier Factor (LOF) anomaly detection classifier, using **Hadamard-test quantum circuits** to estimate inner products between vectors, which are then used to calculate pairwise distances for LOF scoring.

* ✅ **Quantum LOF step** using Qiskit 2.0.2 compatible circuits
* ✅ **LOF score computation**
* ✅ Dual downstream models for clean and noisy regions
* ⚠️ Currently lacks Grover-based quantum minimum/average steps

---

## 🧠 Algorithm Mapping to Paper (Guo 2023 - arXiv:2304.08710)

| Paper Section          | Functionality                           | Implemented? | Notes                                                                                                                       |
|------------------------|-----------------------------------------|--------------|-----------------------------------------------------------------------------------------------------------------------------|
| III-A Eq.(7)           | Amplitude embedding                    | ✅ Yes      | Implemented via `sin²` encoding: `amplitudes = sin(x_i)`.                                                              |
| III-A Eq.(11)          | Swap test ⟨x｜y⟩                   | ✅ Yes      | Implemented manually using CSWAP gates and `initialize()`.                                                                 |
| III-A Eq.(15) - Eq.(17)| Distance from inner product            | ⚠ Partially | Uses `d(x, y) = √(1 - ⟨x｜y⟩)` assuming normalized inputs (not amplitude-based overlap fidelity).                      |
| III-A Step 1.6–1.7      | Quantum Minimum Search                | ❌ No       | Classical sort is used instead of quantum minimum finding.                                                                  |
| III-B                  | Quantum LRD (inverse of avg reach dist) | ❌ No       | Classical averaging is used for local reachability density calculation.                                                    |
| Eq.(2), Eq.(28)        | Grover-based anomaly extraction        | ❌ No       | Anomalies are detected classically using a threshold on LOF score (delta).                                                 |

---

## 🚀 Installation

```bash
pip install quantlof
```

---

## 🧪 Example Usage

```python
from quantum_lof import QuantumLOFClassifier

clf = QuantumLOFClassifier(
    n_neighbors=20,
    delta=1.5,
    quantum_backend='qiskit_simulator',  # or actual IBM backend like 'ibm_cairo'
    shots=1024,
    random_state=42
)

clf.detect_anomalies(X, y)
print(clf.lof_scores_)

anom_idx = clf.get_anomaly_indices()
clean_idx = clf.get_clean_indices()
print("Anomalies indices:", anom_idx)
print("Clean indices:", clean_idx)
```

---

## ⚙️ Core Features

* ✅ Hadamard-test for ⟨x|y⟩ inner products
* ✅ Euclidean distance via Eq. (15–17)
* ✅ k-distance via quantum estimation
* ✅ Local Reachability Density (LRD)
* ✅ LOF scores (Eq. 18) with thresholding
* ✅ Clean/noise classification downstream
* ✅ Fallback to classical when `n > maxsample_for_quantum`

---

## 🛠️ API

### `QuantumLOFClassifier(...)`

| Argument                | Description                                               |
| ----------------------- | --------------------------------------------------------- |
| `n_neighbors`           | Number of neighbors for LOF                               |
| `delta`                 | LOF threshold (LOF ≥ δ → anomaly)                         |
| `quantum_backend`       | Qiskit backend (e.g. `"qiskit_simulator"`, `"ibm_cairo"`) |
| `shots`                 | Number of shots in Hadamard test                          |
| `maxsample_for_quantum` | Fallback threshold for classical mode                     |
| `clean_model`           | Classifier for clean samples (default: SVM)               |
| `noise_model`           | Classifier for all samples (default: RandomForest)        |

---

## 🤖 Implementation Notes

* The quantum inner product estimation uses **Hadamard test** with controlled inverse `StatePreparation`.
* Classical LOF score is maintained to allow hybrid quantum–classical behavior.
* The quantum part only replaces pairwise distance calculations.

---

## 📜 License

MIT License © 2025 [Yuto Mueller](mailto:geoyuto@gmail.com)

---

## 💡 Future Work

* Grover-based minimum search
* QRAM emulation
* Amplitude estimation for LOF
* GPU-accelerated classical fallback
