Academia.eduAcademia.edu

Outline

Title
Abstract
Figures
Background to Quantum Neural Networks
Experimental Investigations
Method
Summary of Formal Language Results
Discussion
Conclusion
References
All Topics
Physics
Quantum Mechanics

Quantum artificial neural network architectures and components

Profile image of Tamaryn MenneerTamaryn Menneer

Information Sciences

https://doi.org/10.1016/S0020-0255(00)00055-4
visibility

…

description

25 pages

link

1 file

Abstract

It is shown by classical simulation and experimentation that quantum artificial neural networks (QUANNs) are more efficient and in some cases more powerful than classical artificial neural networks (CLANNs) for a variety of experimental tasks. This effect is particularly noticeable with larger and more complex domains. The gain in efficiency is achieved with no generalisation loss in most cases. QUANNs are also more powerful than CLANNs, again for some of the tasks examined, in terms of what the network can learn. What is more, it appears that not all components of a QUANN architecture need to to be quantum for these advantages to surface. It is demonstrated that a fully quantum neural network has no advantage over a partly quantum network and may in fact produce worse results. Overall, this work provides a first insight into the expected behaviour of individual components of QUANNs, if and when quantum hardware is ever built, and raises questions about the interface between quantum...

Figures (13)
A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255 Fig. 1. A diagram of a modern version Young’s double-slit experiment (see main text). The peaks on the detector screen when both slits are open (extreme right of diagram) represent the arrival of particles — each of which travels in the form of two waves (one from each slit) — either in phase or out of phase.
A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255 Fig. 1. A diagram of a modern version Young’s double-slit experiment (see main text). The peaks on the detector screen when both slits are open (extreme right of diagram) represent the arrival of particles — each of which travels in the form of two waves (one from each slit) — either in phase or out of phase.
A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255 Fig. 2. The pattern of arrival of individual photons at the detector screen with either one or both slits open can be explained in quantum mechanical terms through an appeal to ‘wave-—particle’ duality: the particle travels as a wave between the gun and detector screen and passes through one or both slits as a wave, but collapses to a particle (point) when the detection screen is encountered. The pattern of arrival now reflects whether the particle arrived at the screen at the peak of its wave (in phase) or at its trough (out of phase). Additionally, the phase of the waves can be shifted if appropriate ‘phase-shifters’ are introduced between slits and screen.
A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255 Fig. 2. The pattern of arrival of individual photons at the detector screen with either one or both slits open can be explained in quantum mechanical terms through an appeal to ‘wave-—particle’ duality: the particle travels as a wave between the gun and detector screen and passes through one or both slits as a wave, but collapses to a particle (point) when the detection screen is encountered. The pattern of arrival now reflects whether the particle arrived at the screen at the peak of its wave (in phase) or at its trough (out of phase). Additionally, the phase of the waves can be shifted if appropriate ‘phase-shifters’ are introduced between slits and screen.
A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
Fig. 3. A quantum neural network can be modelled on the double slit experiment. The photon gun now becomes the input pattern, the slits the input neurons, the waves between the slits and the screen the connections between the input and output neurons, and the screen the output neurons. This differs from the Chrisley approach in that, here, the inputs and weights are kept distinct. Chrisley originally proposed that the slits could be weights themselves. However, in our approach, it is possible to amend the weights (through forms of phase-shifting) independently of modifying the slits. Such phase-shifting can be used to implement back-propagation, for instance (see Fig. 2).
Fig. 3. A quantum neural network can be modelled on the double slit experiment. The photon gun now becomes the input pattern, the slits the input neurons, the waves between the slits and the screen the connections between the input and output neurons, and the screen the output neurons. This differs from the Chrisley approach in that, here, the inputs and weights are kept distinct. Chrisley originally proposed that the slits could be weights themselves. However, in our approach, it is possible to amend the weights (through forms of phase-shifting) independently of modifying the slits. Such phase-shifting can be used to implement back-propagation, for instance (see Fig. 2).
Each type of network will be designated by [x1 x2 x3 x4] So, for instance, [qcoqe] = quantum input to hidden unit weights with classical hidden to output unit weights with quantum hidden unit activations with classical output unit activations
Each type of network will be designated by [x1 x2 x3 x4] So, for instance, [qcoqe] = quantum input to hidden unit weights with classical hidden to output unit weights with quantum hidden unit activations with classical output unit activations
Fig. 5. The 16 different types of network for a classical 2:2:1 (two input units, two hidden units, one output unit) formed from using combinations of (c)lassical and (q)uantum architectures for the input-to-hidden, hidden-to-output, hidden units and output units sections. The network types represented are for two training patterns of two bits each, hence two input neurons. The classical architecture (top left) consists of two input neurons, two hidden units and one output neuron. Quantum components involving connections are subsequently depicted by multiple connections between layers, and quantum components involving units by multiple nodes within a layer. A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
Fig. 5. The 16 different types of network for a classical 2:2:1 (two input units, two hidden units, one output unit) formed from using combinations of (c)lassical and (q)uantum architectures for the input-to-hidden, hidden-to-output, hidden units and output units sections. The network types represented are for two training patterns of two bits each, hence two input neurons. The classical architecture (top left) consists of two input neurons, two hidden units and one output neuron. Quantum components involving connections are subsequently depicted by multiple connections between layers, and quantum components involving units by multiple nodes within a layer. A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
The animate—inanimate classification domain* * This contains the best QUANN results, given by QUANN-type ceqe, together with the results of statistical tests. A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
The animate—inanimate classification domain* * This contains the best QUANN results, given by QUANN-type ceqe, together with the results of statistical tests. A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
The mushroom classification domain® * This contains the best QUANN results, given by QUANN-type eeqe, together with the results of statistical tests.
The mushroom classification domain® * This contains the best QUANN results, given by QUANN-type eeqe, together with the results of statistical tests.
“See the main text for an explanation of the terms. Regular languages domains*
“See the main text for an explanation of the terms. Regular languages domains*
Context-free and type-zero languages domains A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
Context-free and type-zero languages domains A. Narayanan, T. Menneer | Information Sciences 128 (2000) 231-255
esults of statistical tests for the formal language domains and the percentage level of significance* * The mean results from Tables 3 and 4 are also included. Comparisons were made between those variations in which the CLANN and the QUANN gave their highest generalisation performance. The first result is for the training time and the second is for the generalisation. The QUANN results are for cccq-type unless otherwise stated. The number of hidden units (HU) is given for those experiments in which a varying number of hidden units was used.
esults of statistical tests for the formal language domains and the percentage level of significance* * The mean results from Tables 3 and 4 are also included. Comparisons were made between those variations in which the CLANN and the QUANN gave their highest generalisation performance. The first result is for the training time and the second is for the generalisation. The QUANN results are for cccq-type unless otherwise stated. The number of hidden units (HU) is given for those experiments in which a varying number of hidden units was used.
A comparison of the QUANN and control results for the language domains* “The highest generalisation of the QUANN, across variations, is given as a percentage of the highest generalisation achieved by the control. The QUANN training time for these variations is also given as a percentage of the control training time. A code is given after each value to indicate whether the difference between the control and the QUANN result was significant (s) or not (n); ‘—’ indicates no test was carried out. All values are for QUANN-type cceq.
A comparison of the QUANN and control results for the language domains* “The highest generalisation of the QUANN, across variations, is given as a percentage of the highest generalisation achieved by the control. The QUANN training time for these variations is also given as a percentage of the control training time. A code is given after each value to indicate whether the difference between the control and the QUANN result was significant (s) or not (n); ‘—’ indicates no test was carried out. All values are for QUANN-type cceq.
Fig. 7. The two best QUANN types, as evidenced by the simulations. alise in such circumstances. Another implication for ANN research of QUANN research is that it has explored methods for quantum implementation of ANNs. It is essential that ools such as ANNs are able to exploit the immense increase in efficiency of hese quantum computers over those currently used. In order to do this it is necessary for ANNs to be implementable on quantum computers by using methods based on quantum theoretical concepts in their operation. There were wo types of QUANN that gave the best generalisation. These were ceqe for the microfeatural domains and cecq for the formal language domains (see Fig. 7). This shows that different QUANN-types lend themselves to different classes of problem. These experiments have provided a first indication as to which QUANN-type to use for a given type of problem. More interestingly and perhaps controversially, the results provided here indicate that it is not nec- essary to have a fully quantum network to achieve improvements over their classical counterparts. This has several advantages. First, we do not have to hrow away everything we have so far learned about neural networks. We can keep the best classical components and add on the quantum components as necessary. That means that we do not have to generate a whole new theory of QUANNs. Instead, we generate new theories of individual quantum compo- nents and graft these onto existing (classical) knowledge of how ANNs work. That does not mean that, ultimately, some new total theory of QUANNs may not be developed which subsumes classical interpretations and theories of CLANNSs. However, our research indicates that it is not necessary to have an understanding of this whole theory of QUANNSs before experimenting with QUANNs with only one quantum component. Secondly, given the widespread scepticism that quantum hardware will ever
Fig. 7. The two best QUANN types, as evidenced by the simulations. alise in such circumstances. Another implication for ANN research of QUANN research is that it has explored methods for quantum implementation of ANNs. It is essential that ools such as ANNs are able to exploit the immense increase in efficiency of hese quantum computers over those currently used. In order to do this it is necessary for ANNs to be implementable on quantum computers by using methods based on quantum theoretical concepts in their operation. There were wo types of QUANN that gave the best generalisation. These were ceqe for the microfeatural domains and cecq for the formal language domains (see Fig. 7). This shows that different QUANN-types lend themselves to different classes of problem. These experiments have provided a first indication as to which QUANN-type to use for a given type of problem. More interestingly and perhaps controversially, the results provided here indicate that it is not nec- essary to have a fully quantum network to achieve improvements over their classical counterparts. This has several advantages. First, we do not have to hrow away everything we have so far learned about neural networks. We can keep the best classical components and add on the quantum components as necessary. That means that we do not have to generate a whole new theory of QUANNs. Instead, we generate new theories of individual quantum compo- nents and graft these onto existing (classical) knowledge of how ANNs work. That does not mean that, ultimately, some new total theory of QUANNs may not be developed which subsumes classical interpretations and theories of CLANNSs. However, our research indicates that it is not necessary to have an understanding of this whole theory of QUANNSs before experimenting with QUANNs with only one quantum component. Secondly, given the widespread scepticism that quantum hardware will ever

Related papers

Quantum Neural Networks: A Novel Architecture for Hybrid Quantum-Classical Algorithms
Dr. Srinivasa Rao Gundu

Research Square (Research Square), 2024

The capacity of quantum computing to tackle complex problems faster than general computers might lead to industry revolutions. However, actual implementation is problematic due to limited qubit coherence and inherent noise. Using Quantum Neural Networks (QNNs), hybrid quantum-classical algorithms successfully address optimization problems by combining the bene ts of both quantum and traditional computer paradigms. The interface layer, the classical layer, and the quantum layer make up the three fundamental parts of the suggested design. The architecture's performance is contrasted with existing methods to demonstrate its advantages in terms of speed, accuracy, and scalability. With the help of this innovative design, di cult issues that are outside the capabilities of conventional computers can now be tackled, offering a workable solution for issues with nance, logistics, and medication development.

View PDFchevron_right
A Tutorial on Quantum Convolutional Neural Networks (QCNN)
Bafrizal Achyard

2020 International Conference on Information and Communication Technology Convergence (ICTC), 2020

Convolutional Neural Network (CNN) is a popular model in computer vision and has the advantage of making good use of the correlation information of data. However, CNN is challenging to learn efficiently if the given dimension of data or model becomes too large. Quantum Convolutional Neural Network (QCNN) provides a new solution to a problem to solve with CNN using a quantum computing environment, or a direction to improve the performance of an existing learning model. The first study to be introduced proposes a model to effectively solve the classification problem in quantum physics and chemistry by applying the structure of CNN to the quantum computing environment. The research also proposes the model that can be calculated with O(log(n)) depth using Multi-scale Entanglement Renormalization Ansatz (MERA). The second study introduces a method to improve the model's performance by adding a layer using quantum computing to the CNN learning model used in the existing computer vision. This model can also be used in small quantum computers, and a hybrid learning model can be designed by adding a quantum convolution layer to the CNN model or replacing it with a convolution layer. This paper also verifies whether the QCNN model is capable of efficient learning compared to CNN through training using the MNIST dataset through the TensorFlow Quantum platform.

View PDFchevron_right
Quantum perceptron over a field and neural network architecture selection in a quantum computer
Teresa Ludermir

Neural Networks, 2016

In this work, we propose a quantum neural network named quantum perceptron over a field (QPF). Quantum computers are not yet a reality and the models and algorithms proposed in this work cannot be simulated in actual (or classical) computers. QPF is a direct generalization of a classical perceptron and solves some drawbacks found in previous models of quantum perceptrons. We also present a learning algorithm named Superposition based Architecture Learning algorithm (SAL) that optimizes the neural network weights and architectures. SAL searches for the best architecture in a finite set of neural network architectures with linear time over the number of patterns in the training set. SAL is the first learning algorithm to determine neural network architectures in polynomial time. This speedup is obtained by the use of quantum parallelism and a non-linear quantum operator.

View PDFchevron_right
Benchmarking Neural Networks For Quantum Computations
Elizabeth Behrman

IEEE Transactions on Neural Networks and Learning Systems, 2019

The power of quantum computers is still somewhat speculative. While they are certainly faster than classical ones at some tasks, the class of problems they can efficiently solve has not been mapped definitively onto known classical complexity theory. This means that we do not know for which calculations there will be a "quantum advantage," once an algorithm is found. One way to answer the question is to find those algorithms, but finding truly quantum algorithms turns out to be very difficult. In previous work over the past three decades we have pursued the idea of using techniques of machine learning to develop algorithms for quantum computing. Here we compare the performance of standard real-and complex-valued classical neural networks with that of one of our models for a quantum neural network, on both classical problems and on an archetypal quantum problem: the computation of an entanglement witness. The quantum network is shown to need far fewer epochs and a much smaller network to achieve comparable or better results.

View PDFchevron_right
Exploration of Quantum Neural Architecture by Mixing Quantum Neuron Designs
Shanglin Zhou

2021

With the constant increase of the number of quantum bits (qubits) in the actual quantum computers, implementing and accelerating the prevalent deep learning on quantum computers are becoming possible. Along with this trend, there emerge quantum neural architectures based on different designs of quantum neurons. A fundamental question in quantum deep learning arises: what is the best quantum neural architecture? Inspired by the design of neural architectures for classical computing which typically employs multiple types of neurons, this paper makes the very first attempt to mix quantum neuron designs to build quantum neural architectures. We observe that the existing quantum neuron designs may be quite different but complementary, such as neurons from variation quantum circuits (VQC) and Quantumflow. More specifically, VQC can apply realvalued weights but suffer from being extended to multiple layers, while QuantumFlow can build a multi-layer network efficiently, but is limited to us...

View PDFchevron_right
Simulations of quantum neural networks
Elizabeth Behrman

Information Sciences, 2000

We explore by simulation ways in which an array of quantum dot molecules could serve as a quantum neural computer. First, we show that a single quantum dot molecule evolving in real time can act as a recurrent temporal quantum neural network. Inputs are prepared by ®xing the initial states of a quantum dot molecule, and outputs determined by reading its value at a given time T later. The nodes of the network are the instantaneous states of the molecule at successive time slices. The nodes interact indirectly through their mutual interaction with local and phononic modes of the substrate. These modes can be preferentially excited optically, and, therefore, controlled externally. The number of excitations can thus be used as trainable``weight'' parameters for a neural network. This network is shown to perform classical logic gates. By preparing the input state as a superposition state, multiple inputs can be encoded as a single initial state. Second, we simulate the possibility of a spatial, rather than temporal, design, as a Hop®eld net. The network consists of a regular array of quantum dot molecules on a suitable substrate. The molecules interact indirectly as before, and, now, with each other directly through Coulombic interactions. Both of the quantum networks have none of the``wiring problems'' of traditional neural nets: the necessary connections are supplied by the physical system itself. Computation is performed by the intrinsic physics of the physical system. The long range character of the phononic interactions takes the net beyond traditional local connectionist structures. The hypothesized increase in

View PDFchevron_right
Evaluation of Different Ansatze Designs for Quantum Neural Network Binary Classifiers
Maha Metawei

2022

Research Square PDF Research Article Evaluation of Different Ansatze Designs for Quantum Neural Network Binary Classifiers Maha A. Metawei, Hazem Said, Mohamed Taher, Hesham ElDeeb, Salwa M. Nassar This is a preprint; it has not been peer reviewed by a journal. https://doi.org/10.21203/rs.3.rs-1919180/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Quantum Machine Intelligence Version 1 posted 17 Oct, 2022 5 You are reading this latest preprint version Abstract For classification problems, ansatz circuits are an efficient and promising quantum-classical machine learning technique. However, a difficult design challenge arises: which quantum circuit should be used for model training? Low depth ansatz circuits have been used in a variety of applications in recent years. The earlier works defined several quantum circuit descriptors to evaluate an ansatz design such as the circuit expressibility, the efficiency with which a quantum circuit may exploit the Hilbert Space, and the entanglement capability, a circuit's capability of detecting data correlation. Additionally in this work, we investigate, measure, and analyze the effect of the number of ansatz repetition layers L, and the expressibility power of the data encoding module. The proposed quantum neural network classifiers were trained on four different classical datasets using 19 hardware-efficient parametrized circuits. The heuristic experimental results show that ansatz designs with high to moderate expressibiliy values outperformed ones with low expressibility values. By analyzing the circuit cost, we see a tendency toward maximizing the number of parameters which helps in the QNN model's convergence. On simulated hardware, we observe that the proposed quantum neural network achieves up to 97.5% prediction accuracy using the angle data encoding module.

View PDFchevron_right
Operations on Quantum Physical Artificial Neural Structures
Florian Neukart

Procedia Engineering, 2014

The advances that have been achieved in quantum computer science to date, slowly but steadily find their way into the field of artificial intelligence. Specifically the computational capacity given by quantum parallelism, resulting from the quantum linear superposition of quantum physical systems, as well as the entanglement of quantum bits seem to be promising for the implementation of quantum artificial neural networks. Within this elaboration, the required information processing from bit-level up to the computational neuroscience-level is explained in detail, based on the combined research in the fields of quantum physics and artificial neural systems.

View PDFchevron_right
Quantum Convolution Neural Network
Petro Chynnyk

Electronics and Control Systems

In this work, quantum convolutional neural networks are considered in the task of recognizing handwritten digits. A proprietary quantum scheme for the convolutional layer of a quantum convolutional neural network is proposed. A proprietary quantum scheme for the pooling layer of a quantum convolutional neural network is proposed. The results of learning quantum convolutional neural networks are analyzed. The built models were compared and the best one was selected based on the accuracy, recall, precision and f1-score metrics. A comparative analysis was made with the classic convolutional neural network based on accuracy, recall, precision and f1-score metrics. The object of the study is the task of recognizing numbers. The subject of research is convolutional neural network, quantum convolutional neural network. The result of this work can be applied in the further research of quantum computing in the tasks of artificial intelligence.

View PDFchevron_right
Hybrid quantum-classical convolutional neural networks
楚 郭

Science China Physics, Mechanics & Astronomy, 2021

Deep learning has been shown to be able to recognize data patterns better than humans in specific circumstances or contexts. In parallel, quantum computing has demonstrated to be able to output complex wave functions with a few number of gate operations, which could generate distributions that are hard for a classical computer to produce. Here we propose a hybrid quantum-classical convolutional neural network (QCCNN), inspired by convolutional neural networks (CNNs) but adapted to quantum computing to enhance the feature mapping process. QCCNN is friendly to currently noisy intermediate-scale quantum computers, in terms of both number of qubits as well as circuit's depths, while retaining important features of classical CNN, such as nonlinearity and scalability. We also present a framework to automatically compute the gradients of hybrid quantum-classical loss functions which could be directly applied to other hybrid quantum-classical algorithms. We demonstrate the potential of this architecture by applying it to a Tetris dataset, and show that QCCNN can accomplish classification tasks with learning accuracy surpassing that of classical CNN with the same structure.

View PDFchevron_right

References (19)

  1. G.A. Carpenter, S. Grossberg, J.H. Reynolds, ARTMAP-supervised real-time learning and classi®cation of nonstationary data by a self-organizing neural network, Neural Networks 4 (5) (1991) 565±588.
  2. R.L. Chrisley, Quantum learning, in: P. Pylkk anen, P. Pylkk o (Eds.), New Directions in Cognitive Science, Proceedings of the International Symposium, Saariselk a, Finnish Arti®cial Intelligence Society, Lapland, Finland, 1995.
  3. R.L. Chrisley, Learning in non-superpositional quantum neurocomputers, in: P. Pylkk anen, P. Pylkk o (Eds.), Brain Mind and Physics, IOS Press, Amsterdam, 1997, pp. 126±139.
  4. H. Everett, Relative state formulation of quantum mechanics, Reviews of Modern Physics 29 (1957) 454±462.
  5. A. Ezhov, Spurious memory, single-class and quantum neural networks, in: Proceedings of the Fifth Joint Conference on Information Sciences, vol. 1, 2000, pp. 767±770.
  6. L.K. Grover, Quantum computers can search arbitrarily large databases by a single query, Phys. Rev. Lett. 79 (1997) 4709±4712.
  7. L.K. Grover, Quantum mechanics helps in searching for a needle in a haystack, LANL Physics e-print archive: http://xxx.lanl.gov/quant-ph/9605043, Bell Labs, Murray Hill, NJ, 1997b, also in Physical Review Letters 79 (1997) 325±328.
  8. S.C. Kak, Quantum neural computing, Advances in Imaging and Electron Physics 94 (1995) 259±314.
  9. M. Lewestein, Quantum perceptrons, Journal of Modern Optics 41 (12) (1994) 2491±2501.
  10. T. Menneer, A. Narayanan, Quantum inspired neural networks, Department of Computer Science, University of Exeter, UK, http://www.dcs.ex.ac.uk/reports/reports.html, 1995.
  11. T. Menneer, A. Narayanan, Quantum arti®cial neural networks vs classical arti®cial neural networks: experiments in simulation, in: Proceedings of the Fifth Joint Conference on Information Sciences, vol. 1, 2000, pp. 757±759.
  12. T.S.I. Menneer, Quantum arti®cial neural networks, Ph.D. Thesis, Department of Computer Science, University of Exeter, Exeter, EX4 4PT, UK, 1998.
  13. A. Narayanan, J. Wallace, Quantum algorithms for route ®nding, in: Proceedings of the Fifteenth European Meeting on Cybernetics and Systems Research, 2000, pp. 140±143.
  14. H.R. Neave, Elementary Statistical Tables, Unwin Hyman Ltd, London, UK, 1981.
  15. P. Pylkk o, Indeterminacy and experience, in: L.F. Niklasson, M.B. Bod en (Eds.), Proceedings of the 1995 Swedish Conference on Connectionism, Lawrence Erlbaum Associates Publishers, Hillsdale, NJ, 1995, pp. 321±329.
  16. C. Robson, Experiment Design and Statistics in Psychology, Penguin, Middlesex, UK, 1973.
  17. D.E. Rumelhart, G.E. Hinton, R.J. Williams, Learning internal representation by error propagation, in: D.E. Rumelhart, J.L. McClelland, PDP Research Group (Eds.), Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. 1, A Bradford Book, MIT Press, Cambridge, MA, 1986.
  18. P. Shor, Algorithms for quantum computation: discrete logarithms and factoring, Proceedings of the 35th Annual Symposium on Foundations of Computer Science, 1994, pp. 124±134.
  19. S. Siegel, Nonparametric Statistics for the Behavioural Sciences, II ed., McGraw-Hill, New York, 1988.

Related papers

"Overview of Quantum Computing in Quantum Neural Network and Artificial Intelligence"
Dr. K A R U N A M U R T H Y . A

“Quing: International Journal of Innovative Research in Science and Engineering, 2023

In recent years, quantum computing has emerged as a potentially gamechanging technology, with applications across various disciplines, including AI and machine learning. In recent years, the combination of quantum computing and neural networks has led to the development of quantum neural networks (QNNs). This paper explores the potential of QNNs and their applications in solving complex problems that are challenging for classical neural networks. This paper explores the fundamental principles of quantum computing, the architecture of QNNs, and their advantages over classical neural networks. Furthermore, this will highlight key research areas and challenges in the development and utilization of QNNs. Through an in-depth analysis, it demonstrates the QNNs hold significant promise for addressing complex computational problems and advancing the field of artificial intelligence.

View PDFchevron_right
New quantum neural network designs
Felix Petitzon

2022

Quantum computers promise improving machine learning. We investigated the performance of new quantum neural network designs. Quantum neural networks currently employed rely on a feature map to encode the input into a quantum state. This state is then evolved via a parameterized variational circuit. Finally, a measurement is performed and post-processed on a classical computer to extract the prediction of the quantum model. We develop a new technique, where we merge feature map and variational circuit into a single parameterized circuit and post-process the results using a classical neural network. On a variety of real and generated datasets, we show that the new, combined approach outperforms the separated feature map & variational circuit method. We achieve lower loss, better accuracy, and faster convergence.

View PDFchevron_right
On Quantum Computers and Artificial Neural Networks
Co. SEP

Quantum computer science in combination with paradigms from computational neuroscience, specifically those from the field of artificial neural networks, seems to be promising for providing an outlook on a possible future of artificial intelligence. Within this elaboration, a quantum artificial neural network not only apportioning effects from quantum mechanics simulated on a von Neumann computer is proposed, but indeed for being processed on a quantum computer. Sooner or later quantum computers will replace classical von Neumann machines, which has been the motivation for this research. Although the proposed quantum artificial neural network is a classical feed forward one making use of quantum mechanical effects, it has, according to its novelty and otherness, been dedicated an own paper. Training such can only be simulated on von Neumann machines, which is pretty slow and not practically applicable (but nonetheless required for proofing the theorem), although the latter ones may be used to simulate an environment suitable for quantum computation. This is what has been realized during the SHOCID (Neukart, 2010) project for showing and proofing the advantages of quantum computers for processing artificial neural networks.

View PDFchevron_right
Training a Quantum Neural Network
Daniel Lima

2003

Quantum learning holds great promise for the field of machine intelligence. The most studied quantum learning algorithm is the quantum neural network. Many such models have been proposed, yet none has become a standard. In addition, these models usually leave out many details, often excluding how they intend to train their networks. This paper discusses one approach to the problem and what advantages it would have over classical networks.

View PDFchevron_right
Recent Developments and Applications in Quantum Neural Network: A Review
Snehashish Chakraverty

Archives of Computational Methods in Engineering, 2018

Quantum neural network is a useful tool which has seen more development over the years mainly after twentieth century. Like artificial neural network (ANN), a novel, useful and applicable concept has been proposed recently which is known as quantum neural network (QNN). QNN has been developed combining the basics of ANN with quantum computation paradigm which is superior than the traditional ANN. QNN is being used in computer games, function approximation, handling big data etc. Algorithms of QNN are also used in modelling social networks, associative memory devices, and automated control systems etc. Different models of QNN has been proposed by different researchers throughout the world but systematic study of these models have not been done till date. Moreover, application of QNN may also be seen in some of the related research papers. As such, this paper includes different models which have been developed and further the implement of the same in various applications. In order to understand the powerfulness of QNN, few results and reasons are incorporated to show that these new models are more useful and efficient than traditional ANN.

View PDFchevron_right

Related topics

  • Engineering
  • Quantum Network Theory

    • Cited by

    Evolution of entanglement in quantum neural network
    Tahir Akinci

    IOP Conference Series: Materials Science and Engineering

    This study concerns with the evolution of entanglement in a quantum neural network (QNN) model that is locally in contact with data environments. As a valuable resource, duration of entanglement in quantum systems is extremely important. Therefore, the effect of various initial states on the occurrence or decay of entanglement are examined in the presence of information reservoirs. In this study, central spin model was investigated as a quantum version of neural networks inspired by biological models. The architecture of the model is based on a central spin system with two nodes where the nodes are coupled to independent spin baths. Numerical results show that initial state preparation has a profound effect on the fate of entanglement. The results show that the entanglement lifetime can be adjusted by engineering the reservoir states as well as the initial states of the system of interest. The results can be used to improve the performance of the formation or distribution of entanglement in realistic communication network states.

    View PDFchevron_right
    Classification of power quality disturbances using quantum neural network and DS evidence fusion
    Sudhir Paraskar

    2012

    This paper proposes a hybrid approach combining Wigner-Ville distribution (WVD) with convolutional neural network (CNN) for power quality disturbance (PQD) classification. Firstly, a WVD technique is developed to transfer a 1D voltage disturbance signal into a 2D image file, followed by a CNN model developed for the image classification. Then, the feature maps are extracted automatically from the image file and different patterns are extracted from variables on CNN. A set of synthetic signals, as well as real-world measurement data, are used to test the proposed method. The high classification accuracy of test results is achieved to confirm the effectiveness of the proposed method. Furthermore, the model is simplified and optimized by visualizing the output of convolutional layers. On this basis, one visualizing technique called the class activation map (CAM) is used to identify the location and shape of "hotspots (PQDs)". The effect of incorrect classification of the model is analyzed with the CAM. Therefore, the proposed method is proved to have the capability of providing necessary and accurate information for PQDs, which will then be used to determine the subsequent PQ remedy actions accordingly. INDEX TERMS classification, convolutional neural network (CNN), deep learning, power quality disturbances, power systems, Wigner-Ville distribution (WVD).

    View PDFchevron_right
    Pattern classification using maximally entangled quantum states (MES)
    Manu Pratap Singh

    The European Physical Journal Plus, 2014

    Pattern classifications have been performed by employing the method of Grover's iteration on Bell's MES and Singh-Rajput MES in a two-qubit system and it has been demonstrated that, for any pattern classification, in a two-qubit system the maximally entangled states of Singh-Rajput eigenbasis provide the most suitable choice of search states while, in no case, any of Bell's states is suitable for such pattern classifications. Applying the method of Grover's iterate on three different superpositions in a three-qubit system, it has been shown that the choice of exclusive superposition, as the search state, is the most suitable one for the desired pattern classifications based on Grover's iterative search algorithm. 1 Introduction Richard Feynman examined the role quantum mechanics can play in the development of future computer hardwares and demonstrated [1] that the time evolution of an arbitrary quantum state is intrinsically more powerful computationally than the evolution of a logical classical state. Since then, quantum computing has attracted wide attention and soon became the hot topic of research, especially after Shor's quantum prime factoring algorithm [2] and Grover's random database search algorithm [3] were proposed. Quantum Computer (QC) is quantum information processing. It is a relatively new discipline and not yet completely understood, however, it provides an excellent introduction to many key ideas. Simon [4] demonstrated the power of quantum computation and proved the quadratic reduction of the amount of quantum data required, if quantum states rather than classical states are transmitted, and Vetura [5] demonstrated the potential of the quantum system to exhibit correlations that cannot be accounted classically. There are two main motivations for applying capabilities of quantum computation to neural networks: to compensate the ever-decreasing scale in hardware development and to produce a computational capability not available in classical neural computation. Zak [6] combined classical and quantum neural networks and developed quantum decision maker and Bshouty and Jackson [7] demonstrated the superiority of quantum learning algorithm over the classical one in certain situations like Quantum Hopfield Networks and Quantum Associative Memories (Qu.AM). Quantum entanglement [8] is one of the most interesting features of quantum mechanics and it provides promising and wide applications in quantum information processing such as teleportation [9], dense coding [10, 11], geometric quantum computation [12, 13], quantum neural computing [14-16], universal quantum computing network [17-19] and quantum cryptography [20-22]. Measurement and manipulation of the entangled state of a many-particle system become a far-reaching consequence of quantum information processing. The physically allowed degree of entanglement and mixture is a timely issue given that the entangled mixed states could be advantageous for certain quantum information situations [23]. The simplest non-trivial multi-particle system that can be investigated theoretically, as well as experimentally, consists of two qubits which display many of the paradoxical features of quantum mechanics such as superposition and entanglement. The basis of entanglement is the correlation that can exist between qubits. From the physical point of view, entanglement is still little understood. What makes it very powerful is the fact that since quantum states exist as superposition, these correlations exist in superposition as well and when superposition is destroyed, the proper correlation is somehow communicated between a e-mail:

    View PDFchevron_right
    Quantum Local Binary Pattern for Medical Edge Detection
    Abdelouahab MOUSSAOUI

    Journal of Information Technology Research

    Edge detection is one of the most important operations for extracting the different objects in medical images because it enables delimitation of the various structures present in the image. Most edge detection algorithms are based on the intensity variations in images. Edge detection is especially difficult when the images are textured, and it is essential to consider the texture in edge detection processes. In this article, the authors propose a new procedure to extract the texture from images, called the Quantum Local Binary Pattern (QuLBP). The authors introduce two applications that use QuLBP to detect edges in magnetic resonance images: a cellular automaton (CA) edge detector algorithm and a combination of the QuLBP and the Deriche-Canny algorithm for salt and pepper noise resistance. The proposed approach to extracting texture is designed for and applied to different gray scale image datasets with real and synthetic magnetic resonance imaging (MRI). The experiments demonstrate...

    View PDFchevron_right
    The Cyborg Philharmonic: Synchronizing interactive musical performances between humans and machines
    Sutirtha Chakraborty

    Humanities and Social Sciences Communications, 2021

    Music offers a uniquely abstract way for the expression of human emotions and moods, wherein melodic harmony is achieved through a succinct blend of pitch, rhythm, tempo, texture, and other sonic qualities. The emerging field of “Robotic Musicianship” focuses on developing machine intelligence, in terms of algorithms and cognitive models, to capture the underlying principles of musical perception, composition, and performance. The capability of new-generation robots to manifest music in a human-like artistically expressive manner lies at the intersection of engineering, computers, music, and psychology; promising to offer new forms of creativity, sharing, and interpreting musical impulses. This manuscript explores how real-time collaborations between humans and machines might be achieved by the integration of technological and mathematical models from Synchronization and Learning, with precise configuration for the seamless generation of melody in tandem, towards the vision of human...

    View PDFchevron_right
    580 California St., Suite 400
    San Francisco, CA, 94104
    © 2025 Academia. All rights reserved