This paper proposes a hybrid quantum-classical algorithm that learns a suitable quantum feature map that separates unlabelled data that is originally non linearly separable in the classical space using a Variational quantum feature map and q-means as a subroutine for unsupervised learning. The objective of the Variational circuit is to maximally separate the clusters in the quantum feature Hilbert space. First part of the circuit embeds the classical data into quantum states. Second part performs unsupervised learning on the quantum states in the quantum feature Hilbert space using the q-means quantum circuit. The output of the quantum circuit are characteristic cluster quantum states that represent a superposition of all quantum states belonging to a particular cluster. The final part of the quantum circuit performs measurements on the characteristic cluster quantum states to output the inter-cluster overlap based on fidelity. The output of the complete quantum circuit is used to compute the value of the cost function that is based on the Hilbert-Schmidt distance between the density matrices of the characteristic cluster quantum states. The gradient of the expectation value is used to optimize the parameters of the variational circuit to learn a better quantum feature map.
This paper proposes a hybrid quantum-classical algorithm that learns a suitable quantum feature map that separates unlabelled data that is originally non linearly separable in the classical space using a Variational quantum feature map and q-means as a subroutine for unsupervised learning. The objective of the Variational circuit is to maximally separate the clusters in the quantum feature Hilbert space. First part of the circuit embeds the classical data into quantum states. Second part performs unsupervised learning on the quantum states in the quantum feature Hilbert space using the q-means quantum circuit. The output of the quantum circuit are characteristic cluster quantum states that represent a superposition of all quantum states belonging to a particular cluster. The final part of the quantum circuit performs measurements on the characteristic cluster quantum states to output the inter-cluster overlap based on fidelity. The output of the complete quantum circuit is used to compute the value of the cost function that is based on the Hilbert-Schmidt distance between the density matrices of the characteristic cluster quantum states. The gradient of the expectation value is used to optimize the parameters of the variational circuit to learn a better quantum feature map.
Clustering is one of the most fundamental tasks in machine learning, and the k-means clustering algorithm is perhaps one of the most widely used clustering algorithms. However, it suffers from several limitations, such as sensitivity to centroid initialization, difficulty capturing non-linear structure, and poor performance in high-dimensional spaces. Recent work has proposed improved initialization strategies and quantum-assisted distance computation, but the similarity metric itself has largely remained classical. In this study, we propose a quantum-enhanced variant of k-means that replaces the Euclidean distance with a quantum kernel derived from the inner product between feature-mapped quantum states. Using the Iris dataset, we use multiple quantum feature maps, including entangled SU2 and ZZ circuits, to embed classical data into a higher-dimensional Hilbert space where cluster structures become more separable. We will also be testing using another dataset, namely the breast cancer dataset. Similarity between data points is computed through the inner product between two states. Our results show that this approach achieves improved clustering stability and competitive accuracy compared to the classical algorithm, with the SU2 feature map yielding an accuracy of 88.6 % on the Iris dataset and 91.0 % on the breast cancer dataset, despite operating on NISQ-feasible shallow circuits. These findings suggest that quantum kernels provide a richer similarity landscape than traditional distance metrics, offering a promising path toward more robust unsupervised learning in the NISQ era.
We propose a novel quantum neural network architecture for unsupervised learning of classical and quantum data based on the kernelized version of Kohonen's self-organizing map. The central idea behind our algorithm is to replace the Euclidean distance metric with the fidelity between quantum states to identify the best matching unit from the low-dimensional grid of output neurons in the self-organizing map. The fidelities between the unknown quantum state and the quantum states containing the variational parameters are estimated by computing the transition probability on a quantum computer. The estimated fidelities are in turn used to adjust the variational parameters of the output neurons. Unlike O(N2) circuit evaluations needed in quantum kernel estimation, our algorithm requires O(N) circuit evaluations for N data samples. Analogous to the classical version of the self-organizing map, our algorithm learns a mapping from a high-dimensional Hilbert space to a low-dimensional grid of lattice points while preserving the underlying topology of the Hilbert space. We showcase the effectiveness of our algorithm by constructing a two-dimensional visualization that accurately differentiates between the three distinct species of flowers in Fisher's Iris dataset. In addition, we demonstrate the efficacy of our approach on quantum data by creating a two-dimensional map that preserves the topology of the state space in the Schwinger model and distinguishes between the two separate phases of the model at θ=π.
Abstract Variational Quantum Classifiers are hybrid algorithms with parameterized circuits, that replicate the neural network structure using both quantum and classical resources. Feature map and Ansatz (parameterized circuit) are the major parts of the algorithm for embedding classical data into quantum states and creating a suitable state that minimizes the cost function with the update of parameters, respectively. The number of times the feature map and the Ansatz are applied in the circuit, greatly affects the accuracy. We propose the optimal number of repetitions of the feature map and Ansatz that will help drive the maximum advantage out of VQCs. With the right combination of repetitions, one can fine-tune the model better.
Quantum classifiers provide sophisticated embeddings of input data in Hilbert space promising quantum advantage. The advantage stems from quantum feature maps encoding the inputs into quantum states with variational quantum circuits. A recent work shows how to map discrete features with fewer quantum bits using Quantum Random Access Coding (QRAC), an important primitive to encode binary strings into quantum states. We propose a new method to embed discrete features with trainable quantum circuits by combining QRAC and a recently proposed strategy for training quantum feature map called quantum metric learning. We show that the proposed trainable embedding requires not only as few qubits as QRAC but also overcomes the limitations of QRAC to classify inputs whose classes are based on hard Boolean functions. We numerically demonstrate its use in variational quantum classifiers to achieve better performances in classifying real-world datasets, and thus its possibility to leverage near-term quantum computers for quantum machine learning.
Unsupervised machine learning is one of the main techniques employed in artificial intelligence. We introduce an algorithm for quantum-assisted unsupervised data clustering using the self-organizing feature map, a type of artificial neural network. The complexity of our algorithm scales as O(LN), in comparison to the classical case which scales as O(LMN), where N is the number of samples, M is the number of randomly sampled cluster vectors, and L is the number of the shifts of cluster vectors. We perform a proof-of-concept demonstration of one of the central components on the IBM quantum computer and show that it allows us to reduce the number of calculations in the number of clusters. Our algorithm exhibits exponential decrease in the errors of the distance matrix with the number of runs of the algorithm.
Quantum operations on pure states can be fully represented by unitary matrices. Variational quantum circuits, also known as quantum neural networks, embed data and trainable parameters into gate-based operations and optimize the parameters via gradient descent. The high cost of training and low fidelity of current quantum devices, however, restricts much of quantum machine learning to classical simulation. For few-qubit problems with large datasets, training the matrix elements directly, as is done with weight matrices in classical neural networks, can be faster than decomposing data and parameters into gates. We propose a method that trains matrices directly while maintaining unitarity through a single regularization term added to the loss function. A second training step, circuit alignment, then recovers a gate-based architecture from the resulting soft-unitary. On a five-qubit supervised classification task with 1000 datapoints, this two-step process produces a trained variational circuit in under four minutes, compared to over two hours for direct circuit training, while achieving lower binary cross-entropy loss. In a second experiment, soft-unitaries are embedded in a hybrid quantum-classical network for a reinforcement learning cartpole task, where the hybrid agent outperforms a purely classical baseline of comparable size.