CERN, WIGNER RCP together with the HEPTech Network are organizing the next
Artificial Intelligence
Academia-Industry Matching Event
Artificial Intelligence, Machine Learning
and those supported by Quantum Algorithms
Hybrid Workshop
The aim of this event is to bring together Academic researchers and Industry experts to share ideas, potential applications and fostering collaborations in the newly emerging field of Artificial Intelligence, Machine Learning, Quantum Algorithms and their application.
Confirmed keynote speakers
Our satellite workshop is GPU DAY 2021
Contributions to the conference are welcome.
Abstracts must contain a title shorter than 100 characters, the name and affiliation of the presenter and coauthors, and a maximum of 4000 characters of body text. Images should be sent separately from the text as the abstract will be reprocessed for display on the website.
Talk submission deadline: 2021 October 31.
The call for abstracts is open.
Registration is now open: fill the registration form here.
Participation is free for students, members of academic institutions, research centers, and universities.
The registration fee is 100 or 150 EUR, and 35000 or 50000 HUF if the participant wants to attend only one or both events. Participants can be pay via bank transfer or with a card at the registration desk.
| Account holder | Wigner Research Centre for Physics |
| Tax number | 15327064-2-43 |
| Bank name | Magyar Államkincstár (Hungarian State Treasury) |
| Announcement | {your name}, GPU Day 2021 |
| Account number | 10032000-01731134-00000000 |
| IBAN | HU15 1003 2000 0173 1134 0000 0000 |
| SEPA transfer | |
| SWIFT code 1 | HUSTHUHB |
| non-SEPA transfer | |
| Correspondent bank | Magyar Nemzeti Bank (Hungarian National Bank) |
| SWIFT code 2 | MANEHUHB |
ORGANIZERS:
András Telcs
Antal Jakovácz
Zoltán Zimborás
The talk will present a discussion about how science is changing, results in data sets with an ever longer effective lifetime. Many of the current sky-surveys (and other mid-scale projects) generate data sets at a cost of tens to hundreds of millions of dollars, yet there is no coherent strategy to ensure the survival of these data sets. We will discuss the different factors related to long term data preservation, the technological changes, and the sociological changes in how data is used, including AI applications. We discuss how academic research can remain competitive in this era of extreme agility.
In recent years, near-term noisy intermediate scale quantum (NISQ) computing devices have become available. One of the most promising application areas to leverage such NISQ quantum computer prototypes is quantum machine learning. While quantum neural networks are widely studied for supervised learning, quantum reinforcement learning is still
just an emerging field of this area. To solve a classical continuous control problem, we use a continuous-variable quantum machine learning approach. We introduce proximal policy optimization for photonic variational quantum agents and also study the effect of the data re-uploading. We present performance assessment via empirical study using Strawberry Fields, a photonic simulator Fock backend and a hybrid training framework connected to an OpenAI Gym environment and TensorFlow. For the restricted CartPole problem, the two variations of the photonic policy learning achieve comparable performance levels and a faster convergence than the baseline classical neural network of same number of trainable parameters.
The simulation of the passage of particles through the detectors of High Energy Physics (HEP) experiments is a core component of any physics analysis. However, a detailed and accurate simulation of the detector response using the Geant4 toolkit is a time and CPU consuming process, making it a challenge to gather enough statistics. This is especially problematic for the upcoming High Luminosity LHC upgrade, with more complex events and a much increased trigger rate. In this talk, we discuss novel machine learning techniques that demonstrate efficient and accurate detector simulation and can be used to bridge the HL-LHC resource gap.
Deep learning is widely used in microscopy applications, mainly for processing the generated data, e.g. image segmentation and classification. Real-time applications are less common due to the research focus to develop the best possible method for a given problem and the time it requires to engineer. However, such approaches help in automating monotonous manual processes, they standardize the experiments and can multiply the number of measurements in the same amount of time. In my talk, I show two examples of automated microscopy systems where deep learning methods have a central role. The first project covers the automation of a patch clamp microscope that allows measuring the electrophysiological properties of neurons in human brain tissue slices. We use a contrast-enhancing technique to visualize the sample, and a trained deep learning method detects the healthy neurons in the tissue. Then, the system automatically moves a micropipette to the selected cell and performs the recording. The second project is about a system that automatically selects 3D cell cultures based on their morphological properties and transfers them into another plate. The included deep learning algorithm segments the cell cultures using stereo microscopy. This automates a sample preparation step when working with spheroids. Both of these systems have been validated and used for hundreds of measurements.
One of the central questions of the digital transformation is how Industry 4.0 can be utilised to fulfil all the production system requirements while following state-of-the-art developments. Effective information management is critical for the improvement of manufacturing processes. This research high- lights that ontologies are suitable for manufacturing management and recommends the multilayer-network based interpretation and analysis of ontology- based databases. A wire harness assembly-based case study will serve as an illustrative example to demonstrate how ontology-based modelling can be utilised with network science tools for system analysis.
Our first goal is to provide an overview of describing the production standards using the most critical elements of semantics & syntax. Furthermore, provide the necessary theoretical background to understand the specific semantic models used in ontologies and description methods of a manufacturing sys- tem. The main principles of semantic modelling are the RDF (Resource Description Framework) [1] and OWL (Web Ontology Language) [2]. The RDF Schema provides interoperability between applications that exchange machine- understandable information on the Web. OWL can develop domain-specific schemas and ontologies (so-called meta-models) and represent the meaning of terms in vocabularies and the relationships between such terms. RDF triples can be utilised to extend a graph between unique data instances, collect general data as well as express semantics, attributes, and hierarchies [3].
The second goal is to investigate the defined descriptive and influential factors with data queries and analyses. For data extraction, we applied SPARQL (Structured Protocol and RDF Query Language) queries, which most significant benefit is creating a structured version from the data stored in the ontology or extracting data from RDF, which is an excellent source to manage basic production analysis. However, if a more in-depth investigation is needed or the production ontology has high complexity, the tools of Data Science can provide more accurate solutions.
Because of the outstanding network analysis and visualisation capabilities, it is worth generating labelled multilayer networks from ontology models and graph databases [4, 5]. The MuxViz software application [6, 7] has been used to create multilayer graph representations to explore further analytical possibilities. The methodology of RDF triples how to link data is the same as in the case of bipartite graphs (which are used during the multilayer formation). Studies are published to formalise the bipartite graphs as an intermediate model for RDF with a goal of graph-based notions in querying and storage [8]. Further- more, an RDF database can be simultaneously analysed as layers of a multilayer network [9], providing a solution for Production Flow Analysis.
And finally, as part of the multilayer system analysis the effective serialisation of nodes and clustering of data points can be utilised in a wide range of applications, such as optimal allocation or analysing complex data-sets to discover patterns and internal information in networks.
We created a benchmark model using ontologies, which can serve as an effective methodology and also provide critical messages during the development for industrial and research parties as well. The reproducible industrial case study (based on wire harness assembly manufacturing process) can give guidance on understanding the ontological modeling of a manufacturing process. We performed an assembly line balancing utilising data science tools based on these data and methodology, which contain further potential analysis and development methods. Based on the results, we concluded that detectable clusters and communities within these networks can facilitate the formation of production cells or grouping resources in the process.
References
[1] D. Brickley, R. V. Guha, A. Layman, Resource description framework (rdf) schema specification, W3C (1999).
[2] D. L. McGuinness, F. Van Harmelen, et al., Owl web ontology language overview, W3C recommendation 10 (10) (2004) 2004.
[3] A. Brodt, D. Nicklas, B. Mitschang, Deep integration of spatial query processing into native rdf triple stores, in: Proceedings of the 18th SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2010, pp. 33–42.
[4] T. Ruppert, G. Honti, J. Abonyi, Multilayer network-based production flow analysis, Complexity 1 (2018).
[5] G. Honti, G. Dorgo, J. Abonyi, Network analysis dataset of system dynamics models, Data in brief 27 (2019).
[6] M. De Domenico, M. A. Porter, A. Arenas, Muxviz: a tool for multilayer analysis and visualization of networks, Journal of Complex Networks 3 (2) (2015) 159–176.
[7] M. D. Domenico, Muxviz - the multilayer analysis and visualization platform., https://muxviz.net/ (2020).
[8] J. Hayes, C. Gutierrez, Bipartite graphs as intermediate model for rdf, in: International Semantic Web Conference, Springer, 2004, pp. 47–61.
[9] G. Honti, J. Abonyi, Frequent itemset mining and multi-layer network-based analysis of rdf
databases, Mathematics 9 (4) (2021) 450.
Analysis of signals by means of mathematical transformations proved to be an effective method in various aspects, such as filtering, system identification, feature extraction, classification etc. The most widely used method in transform-domain techniques operates with fixed basic functions like the trigonometric functions in the Fourier transform, Walsh functions in the Walsh–Fourier transform, mother wavelet function for the wavelet transform, etc. In these cases, the flexibility of the method is in the proper choice of the function system. Once the system is set it is used regardless the difference between the individual signals. In other words, the system can be adjusted to the problem level, but not to the individual signal level. This limitation turned to be significant in such cases, especially in dynamically changing environments, where we need to adjust the system to the signal. One way to surpass this limitation is to use adaptive orthogonal transformations. In recent years, we generalized this concept, developed various adaptive mathematical models, and successfully applied them in a range of applications, including ECG, EEG signal processing, telecommunication, CT-, photoacoustic-, and thermographic imaging. In these applications the transformation step was followed by various machine learning techniques. Even though the interaction of these two phases were studied and considered but they were not integrated into a uniform method. In our recent project, by utilizing our former work, we incorporate the representation abilities of adaptive orthogonal transformations and the prediction abilities of neural networks (NNs) in form of a joint model called VPNet. This is a novel model-driven NN architecture for 1D signal-processing problems which utilize variable projection (VP). Applying VP operators to neural networks has the advantage of learnable features, interpretable parameters, and compact network structures. We show that, compared to fully connected and one-dimensional convolutional networks, VPNet offers fast learning ability and good accuracy at a low computational cost of both training and inference. Based on these advantages and the promising results obtained, we anticipate a profound impact on the broader field of signal processing, in particular on classification, regression and clustering problems. In order to demonstrate its efficiency, we evaluated the performance of the VPNet approach in three tasks including classification of normal and abnormal heartbeats in real electrocardiogram (ECG) signals, color classification of visually evoked potentials in EEG signals, and road surface detection based on wheel sensor data.
Our research focuses on the suitability of universal model explanatory tools and methods for qualitative abstractions of embedded AI models for V&V purposes.
The rapidly spreading solutions based on embedded artificial intelligence in cyber-physical systems have defined the behavioral model of complex systems with machine learning tools. A fundamental obstacle to their prevalence is that accurate modeling can only be satisfied with complex models, whose validation phase - especially in critical systems - can be problematic in terms of the interpretability of the model and the explanation of its behavior.
Growing demand is discernable in the field of Artificial Intelligence, with the primary intention of improving the explainability of data sets and derived models (xAI). Various proposed techniques use directly explainable AI model structures, sacrificing some accuracy, or seeking universal tools that derive explanations, independently of the modeling paradigm.
Qualitative modeling plays a special role in the foundations of model-based supervisory control, in which a high-level overview is provided by a discrete model in accordance with the concept of hybrid modeling. The core idea behind the abstraction is mapping entire subsets of the continuous state space corresponding to an operational domain exposing similar behavior to a single qualitative state.
The principle of qualitative modeling has a long tradition in different fields of science as descriptive means; however its use for CPS control raises specific challenges: (i) primarily the quality of the model must be guaranteed, as the decisive factor of its faithfulness; (ii) insufficient coverage of potentially dangerous operational domains may lead to hazardous control errors, (iii) outliers need special care in critical applications.
In our submission, a general overview will be given on the major motivations, intentions, and challenges behind the concept of XAI. So far, many open-source explainable AI libraries are available, including a comprehensive set of algorithms that cover different dimensions of explanations and proxy explainability metrics. Since there is no formal definition of interpretability, it can be approached in various ways, and for each approach, a different library/algorithm is provided. Using two state-of-the-art explainability libraries (IBM AIX360, DALEX), these different approaches will be reviewed via showcasing their effectiveness on the wide scale of machine learning models equipped with different interpretability characteristics.
Our research proposes an approach that allows for qualitative abstraction-level validation of embedded models known even at the black box level. This technique combines dimensionality reduction and clustering methods that accurately separate the operational domains while also recognizing outliers using well-fitting cluster borders. Various interpretability methods will be used for highlighting the cohesive factors amongst the operating regions and guide the understanding of the functionality as well.
Our approach aligns with the current ongoing trend to shift from the extensive domain knowledge, (computationally heavy) construction, and fitting of models towards the V&V and interpretation of the model. We believe that introducing increasing (semi-)automation to the modeling part leads to a better understanding of the analyzed data.
In complex systems, the importance of empirical data analysis-based testing, verification, and validation increases to assure a proper level of service under the typically varying workload. Scaling of these systems needs reusable and scale-independent models for reconfigurability.
The limited faithfulness of speculative analytic models does not support complex system identification. This way, empirical system identification from observations is emerging in this field. The increasing complexity necessitates explainable and well-interpretable models that follow the logic of everyday thinking to validate the model and its use in operation.
This presentation highlights how the empirical model extraction can integrate into the system identification process and presents a qualitative reasoning-based approach for model generalization, consistency checking, model verification using answer set programming.
An interesting research direction is to combine complex and accurate machine learning with the easy interpretability and explainability of qualitative models. New fundamental research from the literature uses Logic Explained Networks (LEN) or Boolean rule generators to create explanations for complex ML models, like neural networks, even if they are black-boxes. Our research combines qualitative modeling with logic interpretation, a well-proven method in qualitative physics to introduce the entire repertoire of discrete formal methods to validate embedded artificial intelligence.