Hyperparameter optimization of hybrid quantum neural networks for car classification

Abstract

Image recognition is one of the primary applications of machine learning algorithms. Nevertheless, machine learning models used in modern image recognition systems consist of millions of parameters that usually require significant computational time to be adjusted. Moreover, adjustment of model hyperparameters leads to additional overhead. Because of this, new developments in machine learning models and hyperparameter optimization techniques are required. This paper presents a quantum-inspired hyperparameter optimization technique and a hybrid quantum-classical machine learning model for supervised learning. We benchmark our hyperparameter optimization method over standard black-box objective functions and observe performance improvements in the form of reduced expected run times and fitness in response to the growth in the size of the search space. We test our approaches in a car image classification task, and demonstrate a full-scale implementation of the hybrid quantum neural network model with the tensor train hyperparameter optimization. Our tests show a qualitative and quantitative advantage over the corresponding standard classical tabular grid search approach used with a deep neural network ResNet34. A classification accuracy of 0.97 was obtained by the hybrid model after 18 iterations, whereas the classical model achieved an accuracy of 0.92 after 75 iterations.

Quantum algorithms applied to satellite mission planning for Earth observation

“…The current work explores the interplay between classical and gate-based quantum computing algorithms. The practical advantage of employing this hybrid approach was shown in our earlier contributions [10]- [12].…”

Quantum Machine Learning: from physics to software engineering

“…Despite the successes outlined in Sec 2.4, the theoretical grounding for such models is limited. We saw that hybrid networks performed well if the quantum section was introduced at the beginning of the model architecture [41] or in the middle [44]. From an information-theoretic perspective, this needs to be investigated in more detail to shed light on the effect of hybridisation.…”

“…In continuation, Ref. [44] suggested a hyperparameter optimisation scheme aimed at architecture selection of hybrid networks. This work also implemented a hybrid network for training, but in two new ways: 1) using a real-world, image recognition dataset [191], and 2) the quantum part of the hybrid network was inserted in the middle of the classical implementation.…”

“…QML is employed in many classical applications. Some notable contributions are in sciences [194][195][196][197][198][199][200], in finance [42,[201][202][203], pharmaceutical [43,45], and automotive industries [44]. In many cases, these models replaced a previously-known classical setting [204][205][206][207].…”

“…Quantum machine learning (QML) is a new paradigm of parallel computation where quantum computing meets network theory for improving computation speed-up and Quantum principal component analysis [27] Quantum support vector machine [28] Quantum algorithm for linear systems of equations [25] Quantum clustering algorithm [26] Hybrid quantum-classical optimization algorithm [29] Variational quantum eigensolver [30] Machine learning creating quantum Quantum approximate machine learning [41] Practical industry applications [42][43][44][45] experiments [31,32] Quantum neural networks [33,34] Quantum error correction with machine learning [35,36] Quantum reinforcement learning [37,38] Graph neural networks detecting quantum advantage [39,40] 2007-2013 2014-2017 2018-2022…”