Quorum : Zero-Training Unsupervised Anomaly Detection using Quantum Autoencoders

Jason Zev Ludmir, Sophia Rebello, Jacob Ruiz, Tirthak Patel · 2025

Read Quorum : Zero-Training Unsupervised Anomaly Detection using Quantum Autoencoders by Jason Zev Ludmir, Sophia Rebello, Jacob Ruiz, Tirthak Patel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-13113.txt.

Introduction

Jason Zev Ludmir, Sophia Rebello, Jacob Ruiz, and Tirthak Patel propose a way to spot odd data in 2025. Their method uses quantum machines and needs no training step. Finding rare, risky events matters in fields such as finance, health, and energy, yet labeled cases are scarce.

The team calls the method Quorum and builds it on quantum autoencoders. Since it learns without labels, it can be used where normal training is hard. Training quantum models is often blocked by hard gradient work, which this design avoids. The paper sets out the method as a first of its kind for this task.

Catalog record

CREATORJason Zev Ludmir, Sophia Rebello, Jacob Ruiz, Tirthak Patel
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-13113

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Quorum : Zero-Training Unsupervised Anomaly Detection using Quantum Autoencoders Jason Zev Ludmir, Sophia Rebello, Jacob Ruiz, Tirthak Patel Abstract Detecting mission-critical anomalous events and data is a crucial challenge across various industries, including finance, healthcare, and energy. Quantum computing has recently emerged as a powerful tool for tackling several machine learning tasks, but training quantum machine learning models remains challenging, particularly due to the difficulty of gradient calculation. The challenge is even greater for anomaly detection, where unsupervised learning methods are essential to ensure practical applicability. To address these issues, we propose Quorum, the first quantum anomaly detection framework designed for unsupervised learning that operates without requiring any training. I Introduction Anomaly detection plays an essential role in various industries, from identifying fraudulent transactions in finance to detecting irregularities in power grids [ 26 , 8 , 25 , 36 ] As datasets grow in complexity, traditional machine learning (ML) methods struggle with scalability and accuracy. Quantum computing offers a promising new approach, with its potential to accelerate computations and detect subtle patterns and correlations in data. Leveraging quantum algorithms for ML tasks, particularly for anomaly detection, could transform how we tackle these challenges [ 28 , 27 , 33 ] . The Challenge. Applying quantum computing to anomaly detection presents significant hurdles. Quantum machine learning (QML) models typically rely on parameterized circuits that require training, which is challenging due to the complexity of computing gradients in quantum systems [ 16 , 31 ] . Quantum systems require gradient calculations from first principles using the parameter shift rule, and these gradients are prone to exponential vanishing in “barren plateau” regions [ 13 ] . Moreover, anomaly detection, by nature, is unsupervised, adding another layer of difficulty since no labeled data is available to guide the training [ 29 , 15 ] . These two factors – quantum training complexity and unsupervised learning requirements – create a considerable challenge for developing efficient quantum anomaly detection methods [ 29 , 14 ] . The Gap. Current quantum-based approaches for anomaly detection fall short because they still require training and often rely on supervised or semi-supervised learning. These methods involve optimizing quantum circuits with labeled data, which limits their applicability in real-world scenarios where such data is scarce. This dependency on training creates both computational overheads and reduces the flexibility needed for fully unsupervised tasks [ 16 , 13 , 21 , 12 , 29 , 14 , 1 ] . Our Solution. To address this, we propose Quorum, the first quantum anomaly detection framework that requires zero training and is designed for unsupervised learning. Quorum leverages quantum principles such as amplitude encoding, random quantum transformations, and SWAP tests to identify anomalies without needing any parameter optimization [ 16 , 13 ] . By utilizing quantum correlations and random projections, Quorum detects anomalies based on deviations from the statistical norms of quantum transformations [ 17 , 21 , 13 ] . First, Quorum’s open-source technique carefully distributes the data into buckets based on the likelihood of anomalous events in the dataset before embedding it into quantum states using amplitude encoding [ 12 , 2 ] . Then, Quorum applies random quantum transformations to this data and uses a SWAP test to measure similarly between quantum states [ 13 , 16 ] . Quorum constructs an “embarrassingly parallelizable” ensemble of such random transformations and leverages statistical measures to identify anomalies. This approach is scalable and flexible, allowing for efficient anomaly detection without the computing cost of gradient calculation and training [ 29 , 21 ] . Quorum’s Evaluation. We evaluate Quorum through extensive experiments on various datasets, including medical, industrial, and lexical data. We compare its performance to a state-of-the-art method that uses a quantum neural network (QNN) which relies on training and supervised labels. Our results show that Quorum has a 23% higher average F1 score over the QNN across evaluated datasets. We also provide an ablation study of how Quorum performs with different-sized subsamples (buckets). Quorum consistently identifies subtle anomalies that the state-of-the-art method may overlook, proving to be an effective zero-training quantum solution for unsupervised anomaly detection [ 29 , 21 ] . II Background Before we present the design of Quorum, we first provide some brief but necessary background. Fig. 1: Visual representation of the steps taken by Quorum to detect the anomalies in a given dataset using quantum computing. II-A Quantum Computing The fundamental unit of quantum computation is the qubit, which exists in a superposition of binary states, represented as \ket{\psi}=\alpha\ket{0}+\beta\ket{1} , where \alpha and \beta are complex numbers that correspond to the amplitudes of the basis states \ket{0} and \ket{1} . The probabilities of measuring the qubit in the \ket{0} or \ket{1} state are |\alpha|^{2} and |\beta|^{2} , respectively, and these probabilities must sum to 1: \norm{\alpha}^{2}+\norm{\beta}^{2}=1 . Qubits can also be entangled, meaning their quantum states are correlated in such a way that they cannot be described independently of one another, which is a key property leveraged in quantum computing applications and algorithms to achieve a quantum advantage [ 4 ] . Quantum gates, represented as unitary matrices, are the basic operations applied to qubits. Common single-qubit gates include the parameterized rotation gates, such as the RX, RY, and RZ gates, which rotate a qubit around the x, y, and z axes, respectively. Two-qubit gates, such as the controlled-X (CX) gate, are used to create entanglement between qubits. These single- and two-qubit gates are defined as: RX(\theta)=\begin{pmatrix}\cos{\frac{\theta}{2}}&-i\sin{\frac{\theta}{2}}\\ -i\sin{\frac{\theta}{2}}&\cos{\frac{\theta}{2}}\end{pmatrix},\hskip 9.24994ptRY(\theta)=\begin{pmatrix}\cos{\frac{\theta}{2}}&-\sin{\frac{\theta}{2}}\\ \sin{\frac{\theta}{2}}&\cos{\frac{\theta}{2}}\end{pmatrix}, RZ(\theta)=\begin{pmatrix}e^{-i\theta/2}&0\\ 0&e^{i\theta/2}\end{pmatrix},\hskip 9.24994ptCX=\begin{pmatrix}1&0&0&0\\ 0&1&0&0\\ 0&0&0&1\\ 0&0&1&0\end{pmatrix} Quantum circuits are constructed by applying sequences of these gates to qubits, followed by measurements. Due to the probabilistic nature of quantum measurements, a circuit’s execution yields a probability distribution of possible outcomes, which often requires many repetitions (or shots) to obtain statistically significant results [ 11 ] , [ 6 ] II-B Quantum Encoding and SWAP Test In quantum computing, classical data must be encoded into quantum states. One common method is Amplitude Encoding , [ 35 ] which maps classical data points x_{0},\dots,x_{n-1} to the quantum state: \ket{\psi}=\textstyle\sum_{j=0}^{n-1}x_{j}\ket{j} . This method allows n classical data points to be encoded into \log_{2}n qubits, enabling efficient representation of high-dimensional data. The SWAP Test is a quantum algorithm used to determine how similar two quantum states are [ 7 ] , [ 22 ] . It computes the inner product \bra{\phi}\ket{\psi} between two states \ket{\phi} and \ket{\psi} . If the states are identical, the test returns a high overlap; otherwise, a lower overlap indicates dissimilarity. This is a key tool in quantum anomaly detection, where dissimilarity between two states may indicate an anomalous data point [ 22 ] . II-C Anomaly Detection Anomaly detection involves identifying data points that deviate significantly from the norm [ 34 , 9 ] . In unsupervised anomaly detection, we assume no labeled data is available, and the goal is to detect anomalies based solely on patterns within the data [ 5 , 30 ] . Common classical techniques include clustering and Isolation Forests. Clustering groups data points based on similarity, with anomalies detected as points far from cluster centroids [ 12 , 1 ] . Isolation Forests, a tree-based algorithm, isolates anomalies by recursively splitting

Read the complete text