AIME21

Contribution List

1. Opening

11/11/2021, 14:00

Online lecture

Opening
Dr. Péter Lévai
Director General
WIGNER RCP

19. Welcome

11/11/2021, 14:10

Dr István Szabó
ELKH
Secretary General

15. Considerations and Applications for HPC, Machine Learning and AI technologies for accelerating scientific discovery at next generation research infrastructures.

Prof. Jonathan Taylor (European Spallation Source ERIC)

Online lecture

Research infrastructures (RI) provide access to state-of-the-art instrumentation for the scientific community for example, The Extremely Brilliant Source at the European Synchrotron Radiation Facility, the European Free Electron Laser and the European Spallation Source.
The business case for these facilities is centred on a service model that provides visiting researchers access to cutting...

14. Conversational AI Technology - Business Applications

Nadia Kaloyanova

Online lecture

We are a company from the private sector and we can share our experience on what are the most common business applications of our (and in general) AI Conversational Technology and Natural Language Understanding.

Some of the most common use cases for our technology are:

• Deep Sentiment and Opinion Analysis - extract content from pdf files, web articles and video files, and analyze it to...

3. Hardness of planning and learning with linear function approximation in reinforcement learning

Csaba Szepesvari (DeepMind/University of Alberta)

Online lecture

Markov decision processes (MDPs) is a minimalist framework that is designed to capture the most important aspects of decision making under uncertainty, a problem of major practical interest. Unfortunately, the price of the minimalist approachis that MDPs lack structure and as such planning and learning in MDPs with combinatorial-sized state and action spaces is strongly intractable. An...

13. Hastlayer - Implementing on Xilinx Alveo Accelerator Cards

Erno David (MTA Wigner FK), Zoltán Lehóczky (Lombiq Technologies Ltd.)

Online lecture

Hastlayer (https://hastlayer.com/) by Lombiq Technologies is a .NET software developer-focused, easy to use high-level synthesis tool with the aim of accelerating applications (possible areas include AI/ML, image/video processing, scientific computations…). It converts standard .NET Common Intermediate Language (CIL) bytecode into equivalent Very High Speed Integrated Circuit Hardware...

12. Insights to Hit to Lead

Dr Ivan Stankov (AI4D.io)

Online lecture

The work discussed here is in progress nevertheless it provides insights into good
practices in drug design, discovery, and development (4D). The final goal is to make the
4D flexible and scalable and reduce the total time for bringing a new chemical entity to
the market a few times. The focus of the presentation is on bio-chemistry data, its
discretization, augmentation, and saturation...

20. Life beyond the pixels: single-cell analysis using machine learning and image analysis methods

Dr Péter Horvath (BRC Szeged, Hungary)

Online lecture

In this talk I will give an overview of the computational steps in the analysis of a single cell-based large-scale microscopy experiments using deep learning techniques. First, I will present a novel microscopic image
correction method designed to eliminate illumination and uneven background effects. New single-cell image segmentation
methods will be presented using differential geometry,...

7. Machine Learning in Government Bond Trading

György Ottucsák

Online lecture

Government bond trading is going through unprecedented change today, the role of traditional (mostly voice broking) bond trading model is decreasing due to the electrification of the markets.
The electrification means more reliable and good quality data that give a solid foundation to the application of advanced statistical approaches.
Our aim is to highlight applications in bond trading...

6. Market liquidity, order book dynamics and machine learning

Balázs Meszéna

Online lecture

Trading in financial exchanges is becoming increasingly data driven and algorithmic due to the large amount of detailed data available about the limit order book and the increasing speed competitive nature of trading. I will argue that because of this, algorithmic trading is a great playground for machine learning methods.
In this talk I will discuss what kind of real-life data science...

9. Methods for Interpreting Text Autoencoders

Bálint Gyires-Tóth (Budapest University of Technology and Economics)

Online lecture

Calculating text statistics, like term frequency and similarity is an essential part of processing and modeling symbol sequences, including, for instance, text, source codes, and DNA sequences. However, calculating these statistics may have high computational demand. Thus, even in case of datasets with a moderate size, it can be cumbersome to integrate it into research or commercial product....

5. nucleAIzer: nucleus segmentation with DL & image style transfer

Réka Hollandi (Biological Research Centre)

Online lecture

Cellular analysis based on microscopy images starts with the identification of cells, typically by segmentation. This challenges researchers to construct out-of-the-box solutions that potentially work in various experiments as downstream analysis depends on segmentation reliability. nucleAIzer is a deep learning-based pipeline intended for the efficient and robust instance segmentation of...

4. Playing detective: Dissecting silent failures of Deep Learning models for 3D Point Clouds

Robert-Zsolt Kabai (Continental)

Online lecture

Silent failure effects in Deep Learning models are challenging, to say the least. Image processing applications have developed various tools in order to help explainability and debugging, including visualizign kernel activations, heat maps, and so on. Learning on 3D point clouds is especially challenging as we deal with an unstructured data of a point set defined in 3D metric space. Models...

16. Representation and control: linear or nonlinear dimensionality reduction?

Prof. Sandro Mussa-Ivaldi Mussa-Ivaldi (Northwestern University)

Online lecture

Divide et impera – divide and rule – is an ancient political strategy broadly applicable to complex problems. Theoretical considerations and experimental evidence suggest that this is how the nervous system generates full repertoires of actions by the combination of pre-defined actions or policies. In the framework of geometry, the linear combination of basis vectors spans a vector space....

2. Superfeminine and Supermasculine Connections of the Human Brain

Vince Grolmusz

Online lecture

The anatomical network of the human brain is one of the most complex and most intriguing object of study the humanity may encounter today. We demonstrate an Artificial Intelligence-based method, which is capable of identifying the most relevant brain connections, which define the sex of the subject. The work is based on the largest and the highest quality braingraph collection to date, the...

8. Taming neural networks with TUSLA: Non-convex learning via adaptive stochastic gradient Langevin algorithms

Dr Attila Lovas (Alfréd Rényi Institute of Mathematics)

Online lecture

Artificial neural networks (ANNs) are typically highly nonlinear systems that are finely tuned via the optimization of their associated, non-convex loss functions. Typically, the gradient of any such loss function fails to be dissipative making the use of widely-accepted (stochastic) gradient descent methods problematic. We offer a new learning algorithm based on an appropriately constructed...

18. The first year of the Hungarian Artificial Intelligence National Laboratory (MILAB)

Dr András Benczúr (Scientific director Artificial Intelligence National Laboratory Institute for Computer Science and Control (SZTAKI))

Online lecture

With the goal of strengthening Hungary's position in AI, MILAB is
founded in 2020 as the coordinated national artificial intelligence
umbrella for the collaboration of all major research centers,
universities and large-scale national programs. With the participation
of 75 researchers, we established a project structure of six main areas
during the kickoff meeting on September 2:
- ...

17. The Hungarian AI strategy

Mr László Boa (Technical Coordinator Hungarian National AI Coalition The Hungarian AI strategy 2020-2030)

Online lecture

In this talk, I present the key institutional background and projects of
Hungary's AI strategy. The goal of the Strategy is to harness the
possibilities inherent in technological change and maximize the
opportunities offered by artificial intelligence (AI). The strategy was
prepared with the involvement of over 250 member organizations of
Hungary’s AI Coalition and over a thousand...

11. Two years into the ESS Control System Machine Learning Project

Karin Rathsman (European Spallation Source ERIC)

Online lecture

The European Spallation Source ERIC (ESS) is a joint European organisation committed to building and operating the world's leading facility for research using neutrons. The facility design and construction includes a powerful linear proton accelerator, a helium-cooled tungsten target wheel and two dozen state-of-the-art neutron instruments.

ESS is made up of a large number of diverse...

10. VPNet: Variable Projection Networks

Dr Péter Kovács (Department of Numerical Analysis, Faculty of Informatics, Eötvös Loránd University, Budapest, Hungary )

Online lecture

Analysis of signals by means of mathematical transformations proved to be an effective method in various aspects, including filtering, system identification, feature extraction, classification, etc. The most widely used method in transform-domain techniques operates with fixed basic functions like trigonometric functions in the Fourier transform, Walsh functions in Walsh–Fourier transform,...