We have invited several high-level specialists all over the world to give lectures and share your knowledge with you. Here is a short overview of their background and abstracts.
Prof. Antti Oulasvirta
Human-computer interaction is an area concerned with the design and study of interactive computing systems for human use. Computational sciences play an increasingly important role in HCI, from generative design to adaptive user interfaces. In this lecture, I review the methodological basis of computational interaction in combinatorial optimization, probabilistic inference, and modelling.
Ms. Mariana Romanyshyn
Lecture 1. Introduction to Natural Language Processing
The first lecture will introduce the area of natural language processing:
- we will discuss useful tasks in NLP and brainstorm possible solutions to them
- we will go through stages of an NLP project from domain analysis to production, paying special attention to working with text data
- we will compare different approaches to NLP with a focus on classical NLP and text processing pipelines
Lecture 2. Language as a Sequence
During the second lecture, we will contrast two methods of language processing: language as a bag of words and language as a sequence. Based on the acquired knowledge, we will develop a classifier to solve an error correction task - commonly confused adjectives and adverbs in the English language (e.g., "I slept good or well?"). The lecture will also include the basics of morphological analysis.
Lecture 3. Language as a Tree
In the third lecture, we will learn to build a syntactic parser from scratch. We will discuss the NLP tasks in which syntactic parsing is of utmost importance, will go through popular syntactic parsing algorithms and finally will develop a dependency parser for the Estonian language. We will use the Estonian UD v2.3 treebank for this task. I can't promise high accuracy of the developed solution, but I can promise a lot of fun.
Computational linguist passionate about building natural language processing applications. I have professional experience in syntactic parsing, sentiment analysis, named entity recognition, fact extraction, text anonymization and a few more areas. For the last five years, I have been working on error correction and text improvement algorithms at Grammarly. I care a lot about computational linguistics, constantly look for talented linguists, and spread the word about the field of NLP by participating at AI conferences, collaborating with Ukrainian universities and organizing educational events. My main interest is structural linguistics as a method of formalizing the natural language.
Prof. Adrian Kaehler
In three lectures, we will survey the emerging technologies that are enabling robots and other autonomous systems to learn how to perform complex tasks based on sensory inputs and experimental interaction with the environment. We will begin with a discussion of neural networks and of the emerging technology called Deep Learning. In the second lecture, we will discuss Reinforcement Learning in a computational context, and some of the ways in which Deep Learning can be used in a Reinforcement Learning context. Finally, in lecture three, we will discuss applications of Deep Reinforcement Learning and finish with a brief discussion of some of the interesting open problems in the field.
Dr. Adrian Kaehler is a recognized expert and inventor in numerous advanced technology domains. Throughout his career, his primary focus has been on intellectual and practical leadership for complex technology innovation efforts for both privately or publicly held companies and enterprises, as well as for the many commercial and government institutions he advises. At this time, Adrian is the founder and CEO of Giant.AI, a venture backed robotics start-up headquartered in Silicon Valley.
He holds a Ph.D. degree in theoretical particle physics from Columbia University which, after enrolling in university at the age of 14 and graduating with a bachelor's degree at 18, he received at the age of 25.
His fields of expertise include robotics, deep learning, artificial intelligence, machine learning, physics, electrical engineering, computer algorithms, machine vision, biometrics, computer games, system engineering, human machine interface, numerical programming, and design. He is the author of numerous papers and over 30 patents in these and other subjects, as well as two a best-selling books on computer vision.
Adj. Prof. Stefania Tomasiello
Granular computing and interpretable models
Interpretable models, with their typical properties, such as transparency, intelligibility, comprehensibility, are becoming more and more important in particular for knowledge usage (e.g. in the medical field). Granular Computing has emerged as a framework for more transparent and accurate models, with application to prediction and classification problems.
In this lecture, some preliminaries on fuzzy sets and information granularity will be first given. Then a new definition of information granule will be introduced and some granular architectures presented, including a deep in time scheme, that is a Granular Recurrent Neural Network. Such computational schemes are able to address the main interpretability issues. In particular, they may also be meant interpretable from a classical mathematical perspective, being formally investigated, by proving properties such as convergence.
Session on Autonomous Driving Lab
Anne Jääger, Tambet Matiisen, Dmytro Fishman, Maksym Semikin, and Jan van Gent
This session introduces UT ICS Autonomous Driving Lab that will be launched in the new Delta Centre in January 2020. Anne Jääger will introduce the project organization, the team and the lab setup we are going to have. Tambet Matiisen will talk about the scope of the project and the scientific goals we are aiming for. Then two main research directions will be presented: Dmytro Fishman will talk about the sensors and high-definition maps required for classical self-driving software stack, and Maksym Semikin will discuss an alternative approach - using neural networks for end-to-end control of the vehicle. We will finish with short Autoware demo in simulation, presented by Jan van Gent.
Joint sessions with IEEE Estonia section
I Session: Augmented and Virtual Reality
"Recent developments in use of virtual and augmented reality in Education" - Prof. Margus Pedaste from The Centre for Educational Technology and Pedagogicum, University of Tartu
"Towards Intelligent Immersive Virtual Environments " - Dr. Aleksei Tepljakov from The Centre for Intelligent Systems, TalTech
II Session: Learning Analytics
"Beyond "just" records: Using multilevel learning analytics to predict students at risk" - Dr. Irene-Angelica Chounta from The Centre for Educational Technology, University of Tartu
"AutoThinking: An adaptive computational thinking game" - Dr. Danial Hooshyar from The Centre for Educational Technology, University of Tartu
III Session: Recent developments in IEEE chapters
"Growing scientific communities through hackathons - Insights from a study on expert mentoring" - Dr. Alexander Nolte from The Institute of Computer Science, University of Tartu
"TalTech new IT study programs - the good, the bad and the ugly..." - Prof. Peeter Ellervee from The Centre for Dependable Computing Systems, TalTech