It is not the model's complexity that gives it an advantage, but rather its compatibility with data.

As time series feature unstructured data and unique processing requirements, they do not easily fit conventional approaches to quantum modeling.

This paper therefore presents various approaches to quantum encoding and decoding of time series data, which are key to their quantum analysis. Each approach has the potential to enhance or impede the model’s ability to represent and manipulate data. Each approach also places constraints on the execution of its quantum model and has a major impact on the model performance.

The paper compares the selected approaches and evaluates their performance.

We explored the fundamental principles of parameterisation and optimisation of quantum models

We revealed the secrets of the process to develop complex quantum models - quantum neural nets!

Quantum models can take advantage of the model features absent from classical systems, while hybrid models can inject the model features missing from the pure quantum systems

This presentation describes development aspects of variational quantum autoencoders. As an example, we will discuss the “worst case” scenario of a quantum autoencoder used in denoising of time-series and signals.

The talk highlights the quantum model’s architectural choices, data encoding strategies, ansatz structures and their parameterisation, circuit measurement alternatives and the interpretation of measurement results. Practical issues in training such models will also be covered, to include the selection of an optimiser and its loss function, as well as the execution of a quantum model on simulators, GPU accelerators and QPUs. Methods of model testing will also be explained.

The lack of quantum model transparency hampers the performance of classical optimizers used in QuTSAE development.

This presentation explains the main design decisions in the development of variational quantum time series models and denoising quantum time series autoencoders.

Although we cover a specific type of quantum model, the problems and solutions are generally applicable to many other methods of time series analysis.

The presentation highlights the benefits and weaknesses of alternative approaches to designing a model, its data encoding and decoding, ansatz and its parameters, measurements and their interpretation, and quantum model optimization. Practical issues in training and execution of quantum time series models on simulators, including those that are CPU and GPU based, as well as their deployment on quantum machines, are also explored. All experimental results are evaluated, and the final recommendations are provided for the developers of quantum models focused on time series analysis.

This talk is about how to break good quantum models with good strategies to make them worse.

Quantum machine learning involves development of models, such as quantum neural networks or quantum support vector machines, which commonly consist of parameterised quantum circuits that can be trained using a hybrid quantum-classical process. The process uses a sample of data to guide the quantum model optimisation, which is performed on a classical computer and which is responsible for varying the circuit parameters to reduce the cost function.

Typically, the cost function is a classical computation estimating the difference between the circuit measurements taking place on a quantum machine and the expected values specified by the data sample. Such a hybrid process can be affected by the emergence of barren plateaus - flat areas in the landscape of the cost function, which prevent effectiveness of the quantum model optimisation.

In this short presentation we will explain the main causes of barren plateaus, outline some strategies for dealing with them, and assess the impact of these strategies on the model's capacity to learn.

Quantum computing is a technological phenomenon, which promises great improvements to information processing but which challenges existing computing practices and standards. Quantum computing harnesses the principles of Quantum Mechanics to support computation, often resulting in algorithms capable of solving previously unsolvable problems and reducing information processing time from years to minutes. Compared with traditional computer systems, quantum computing works on entirely different principles, which at this stage of technological development requires scientific knowledge and specialist training, which forces early adopters of this technology, such as financial and manufacturing organisations, to seek the skills of mathematicians, physicists and machine learning researchers.

This seminar provides an overview of quantum computing concepts, its current developments and applications, available software and hardware, as well as its benefits and pitfalls. It will answer the fundamental questions whether or not the new technology is a revolution of computing, a stepwise refinement of the current trends, is it a threat to our future, or perhaps a blessing for human development.

Training of variational quantum models, such as quantum neural networks, can be affected by the emergence of barren plateaus - flat areas in the landscape of the cost function, which negatively impact the effectiveness of the model parameter optimisation. There exist several approaches to dealing with the formation of barren plateaus. In our research, we investigated three most commonly used approaches, i.e. restricting the circuit depth, effective parameter initialisation, and selection of an appropriate cost function. In some cases, such techniques, while avoiding barren plateaus, may actually reduce the model capacity to learn.

This project therefore investigated an impact of barren plateaus countermeasures on the quantum model capacity to learn, which we measured in terms of the model effective dimension, which scales with the model ability to generalise its training data.

Figure. Research methods and visualisation of the effective dimension (Click image)

Quantum computing is a fascinating new area of technological development, which aims at exploiting processes occurring between elementary particles to perform highly efficient computation. Compared with traditional computer systems, quantum computing works on entirely different principles, allowing tackling problems which until recently have been considered practically unsolvable with classical computer hardware.

The new discipline requires expertise intersecting Particle Physics and Data Science, and good knowledge of advanced Mathematical concepts. And yet, quantum computing attracts enthusiasts, researchers, and developers able to create novel solutions to complex problems in science, business, and engineering.

This seminar will provide a bird view of quantum computing concepts, some of its accomplishments and its promises in diverse application domains, including Astronomy, Weather Forecasting, Chemistry, Pharmacology, Finance, Logistics, Optimisation, Computer Communication and Security, Machine Learning, and of course software and hardware for quantum computing.

At the same time, the seminar will also explain the challenges and limitations of the quantum technology.

Figure. Quantum computing application areas (Click image)

This workshop presents the concepts and methods of Quantum Time Series Analysis. While time series analysis and forecasting are essential for AI business applications, this area seems to be neglected by the mainstream researchers and practitioners of Quantum Machine Learning. There are several reasons for this situation - the first is that well-established approaches to classical time series analysis are already computationally efficient (no need for quantum methods), the second is discrepancy between known time series algorithms and quantum computing methods (no simple pathway to quantum solutions), and the third is that dealing with temporal data with quantum technology is difficult (steep learning curve). And yet, there exist examples of successful quantum computing solutions to time-series analysis in Finance, Medicine and Engineering.

This workshop aims to provide the participants with the foundations of quantum time series analysis, its main concepts, examples of applications where quantum methods are likely to give quantum advantage, and the relevant quantum machine learning algorithms.

The workshop is practical and offers opportunities of hands-on exercises in quantum time series analysis, which cover quantum encoding of temporal data, simple manipulation and analysis of such data, variational and stochastic methods of time series fitting and forecasting (e.g. using quantum neural networks). The provided Jupyter notebook will require some knowledge of Python, the basics of Qiskit (and IBM Quantum environment) and the concepts of quantum neural network techniques.

Figure. Quantum encoding of a serial sliding window model for time series analysis (Click image)

In recent years, breakthroughs in quantum computing methods, as supported by the widely available software and hardware, provided great opportunities for advancing quantum machine learning (QML). In some cases, in spite of limitations of the current quantum technology, such efforts lead to great improvement in effectiveness and efficiency of QML algorithms. In other cases, where efficiency gains may not be apparent at this point in time, quantum researchers developed completely new approaches to data analysis, creation of QML models and their applications. Examples of such work include quantum neural networks, quantum support vector machines, quantum kernel methods and quantum optimisation methods, to name just a few.

Surprisingly, in spite of the potential benefits, there is very little research on applying QML methods to time series processing, such as the effective encoding of time-based data on quantum devices, analysis of such representations and interpretation of results obtained from quantum algorithms to support decision making.

This seminar introduces the main QML concepts and their relevance to time series processing. It explains a selection of novel quantum algorithms for time series analysis and demonstrates their workings in IBM Qiskit.

Figure. Quantum encoding of a serial sliding window model with overloading for time series analysis (Click image)

While quantum computing attracts huge business and government investments and its research results are magnified by popular press, the expectations of practical breakthroughs in business applications of this emerging technology are incredibly high. This presentation dispels some of the common myths of quantum computing, explains its reality and some of the promises claimed by quantum technology vendors and researchers.

By walking through the fundamental principles of quantum computing, the currently used tools and practical applications, the seminar clears both the hype and negativity towards the new technology, and casts some light on its current trends, progress, and timelines.

The seminar clarifies some of the challenges faced by quantum hardware manufacturers, the gaps between quantum software functionality vs its utility and productivity, difficulties in the transition of quantum applications from lab to practice, and the severity of skill shortages in the quantum computing market. Finally, software and application development opportunities for this technology are highlighted.

In conclusion, the future of quantum computing is optimistic with huge potential for researchers, vendors, application developers, and investors.

Figure. Organisations which have a significant impact on the development of quantum technology (Click image)

This seminar explains a typical pathway to a career in quantum computing. It discusses the reasons for the recruiters in this area to search for a wide range of knowledge, skills and talents. It warns the newcomers that quantum computing demands some prior study of maths and programming. It explains the knowledge scaffolding including quantum computing tools, the discipline foundations, machine learning and hardware. The seminar then outlines the opportunities for learning and engagement with the quantum community. Finally, it promotes online training courses, regular challenges and hackathons, which give the learners the edge for learning state-of-the-art of quantum computing.

Figure. The workings of a quantum community (Click image)

Quantum computing is a new and rapidly evolving area of research and development.

It concerns building and using information processing systems, which are capable of harnessing phenomena at atomic and sub-atomic scale. Such systems work on entirely new principles of computation and offer breakthroughs in solving problems, which are deemed practically unsolvable with classical machines.

In the last few years, quantum computing captured the attention of computer vendors, research labs and the enterprise. It attracted phenomenal government grants and business investment, and stirred imagination of academics, students, and citizen scientists. And yet, its fundamental concepts are commonly misunderstood, its claims grossly exaggerated and promises misplaced.

This seminar therefore aims to give a high-level overview of quantum computing and its sub-discipline quantum machine learning. It explains their basic concepts, their tools and devices, as well as their practical applications. It then discusses the benefits and limitations of quantum technology. It finally, outlines the future challenges and opportunities for research and education.

Figure. The QML principles (Click image)

Developments in text analytics are enabling computer-assisted analyses of large bodies of research literature. Unsupervised text analytics can yield results beyond what traditional human-directed coding and analysis can produce, by letting the literature speak, rather than imposing pre-conceived conceptual mapping on the process.

This research explores how a large body of data science literature was analysed through unsupervised text analytics.

Based on our experience of this process, advice in the research methods literature, and hermeneutic principles, we are currently developing our learning into a more generic method. The method will identify steps involved in this iterative process, guiding principles, caveats and pitfalls for researchers seeking to utilise machine-assisted analysis of large bodies of published research.

Figure. Extraction and mapping of domain concepts with text analytics (Click image)

The field of data science emerged in recent years, building on advances in computational statistics, machine learning, artificial intelligence, and big data.

Modern organizations are immersed in data and are turning toward data science to address a variety of business problems. While numerous complex problems in science have become solvable through data science, not all scientific solutions are equally applicable to business. Many data-intensive business problems are situated in complex socio-political and behavioral contexts that still elude commonly used scientific methods.

To what extent can such problems be addressed through data science? Does data science have any inherent blind spots in this regard? What types of business problems are likely to be addressed by data science in the near future, which will not, and why?

In this seminar, we develop a conceptual framework to inform the application of data science in business. The framework draws on an extensive review of data science literature across four domains: data, method, interfaces, and cognition. We draw on Ashby’s Law of Requisite Variety as theoretical principle.

We conclude that data-scientific advances across the four domains, in aggregate, could constitute requisite variety for particular types of business problems. This explains why such problems can be fully or only partially addressed, solved, or automated through data science. We distinguish between situations that can be improved due to cross-domain compensatory effects, and problems where data science, at best, only contributes merely to better understanding of complex phenomena.

Figure. Requisite variety framework for applications of data science in business (Click image)