Introduction to Tools and Methods of Artificial Intelligence

Opening & Introduction to AI

Machine Learning and Deep Learning, Lecture 

Natural Language Processing, Lecture 

Natural Language Processing, Lecture 

Dr. János Botzheim

Natabara Gyöngyössy

Natabara Gyöngyössy

András Simonyi

András Simonyi

This course provides students with a comprehensive understanding of the inner workings of Deep Neural Networks and equips them with the necessary skills to implement them while maintaining ethical principles that contribute to the advancement of humanity. The course is designed to achieve the following learning objectives:
• Develop a foundational understanding of Deep Learning
• Gain practical experience in implementing Neural Network architectures
• Become adept at using PyTorch for Deep Learning
• Develop proficiency in the use of open-source Neural Network software
The course content is regularly updated to ensure participants learn the latest techniques in the field. Participants will learn modules that cover a range of topics including Linear Regression, Artificial Neural Networks, Classification, Object Detection, Semantic and Instance, Segmentation, Human Pose Estimation, and Miscellaneous topics.

This course provides students with a comprehensive understanding of the inner workings of Deep Neural Networks and equips them with the necessary skills to implement them while maintaining ethical principles that contribute to the advancement of humanity. The course is designed to achieve the following learning objectives:
• Develop a foundational understanding of Deep Learning
• Gain practical experience in implementing Neural Network architectures
• Become adept at using PyTorch for Deep Learning
• Develop proficiency in the use of open-source Neural Network software
The course content is regularly updated to ensure participants learn the latest techniques in the field. Participants will learn modules that cover a range of topics including Linear Regression, Artificial Neural Networks, Classification, Object Detection, Semantic and Instance, Segmentation, Human Pose Estimation, and Miscellaneous topics.

Our Natural Language Processing (NLP) course aims to give a hands-on overview to our students on the recent advancements in the field, while also providing the needed theoretical foundation. We will revisit the idea of the Transformer Deep Learning architecture and how it turned the classical NLP pipelines into modern generative AI chatbots.

Transformer-based NLP architectures
o Intro to the Transformer architecture
o Language modeling tasks with Transformers
o Using pre-trained language models on downstream tasks

Our Natural Language Processing (NLP) course aims to give a hands-on overview to our students on the recent advancements in the field, while also providing the needed theoretical foundation. We will revisit the idea of the Transformer Deep Learning architecture and how it turned the classical NLP pipelines into modern generative AI chatbots.

Prompt Engineering
o Discrete and continuous prompt engineering methods
o Answer engineering, and combining prompts
o Prompting in “instruct” and “chatbot”-like use-cases
 

Introduction to AI

Machine Learning and Deep Learning, Practice

Machine Learning and Deep Learning, Practice

Natural Language Processing, Practice 

Natural Language Processing, Practice 

Dr. János Botzheim

Szilárd Kovács 

Szilárd Kovács 

This lecture provides the information needed to take the first steps in the field of Artificial Intelligence. Here we provide a brief historical overview of the original ideas and main concepts of AI which are still relevant in state-of-the-art scenarios. The lecture features the main problem types of AI and their solution via state-space encoding and pathfinding. The standard solution methods, graph representations, and rule-based systems are also presented.

This course provides students with a comprehensive understanding of the inner workings of Deep Neural Networks and equips them with the necessary skills to implement them while maintaining ethical principles that contribute to the advancement of humanity. The course is designed to achieve the following learning objectives:
• Develop a foundational understanding of Deep Learning
• Gain practical experience in implementing Neural Network architectures
• Become adept at using PyTorch for Deep Learning
• Develop proficiency in the use of open-source Neural Network software
The course content is regularly updated to ensure participants learn the latest techniques in the field. Participants will learn modules that cover a range of topics including Linear Regression, Artificial Neural Networks, Classification, Object Detection, Semantic and Instance, Segmentation, Human Pose Estimation, and Miscellaneous topics.

This course provides students with a comprehensive understanding of the inner workings of Deep Neural Networks and equips them with the necessary skills to implement them while maintaining ethical principles that contribute to the advancement of humanity. The course is designed to achieve the following learning objectives:
• Develop a foundational understanding of Deep Learning
• Gain practical experience in implementing Neural Network architectures
• Become adept at using PyTorch for Deep Learning
• Develop proficiency in the use of open-source Neural Network software
The course content is regularly updated to ensure participants learn the latest techniques in the field. Participants will learn modules that cover a range of topics including Linear Regression, Artificial Neural Networks, Classification, Object Detection, Semantic and Instance, Segmentation, Human Pose Estimation, and Miscellaneous topics.

Our Natural Language Processing (NLP) course aims to give a hands-on overview to our students on the recent advancements in the field, while also providing the needed theoretical foundation. We will revisit the idea of the Transformer Deep Learning architecture and how it turned the classical NLP pipelines into modern generative AI chatbots.

o Practice: Huggingface Transformers for NLP tasks

Our Natural Language Processing (NLP) course aims to give a hands-on overview to our students on the recent advancements in the field, while also providing the needed theoretical foundation. We will revisit the idea of the Transformer Deep Learning architecture and how it turned the classical NLP pipelines into modern generative AI chatbots.

o Practice: Applied prompt engineering using Mistral and GPT-4

Introduction to ML

Fuzzy systems

Evolutionary Algorithms

Spiking Neural Networks

AI research ethics and legal aspects

Dr. Menyhárd-Balázs Krisztina

Machine Learning is a foundation sub-field of modern AI. Focusing on the concept of learning this lecture aims to provide an introduction to different approaches, such as supervised, unsupervised, or reinforcement learning. The standard architecture of model-problem-solution and the importance of labels are presented. After exploring the main ideas of machine learning the corresponding linear and neural models will be explored such as Linear Regression, Logistic Regression, Perceptron, and
k-Means. Students are also introduced to the fundamental questions of linearity and its relation to the universal approximation theorem.

Boolean logic is the mainstream solution for describing logical relations in computer science. It excels in defining exact rules but fails to grasp the uncertainty and multivaluedness of human cognition. The evaluation of everyday concepts cannot be done with crisp (binary) values, thus the application of Fuzzy Logic is needed. 

Evolutionary algorithms are a class of stochastic heuristic optimization algorithms inspired by evolutionary biology and natural selection. They have been around for well over a century now solving NP problems and remain significant in optimization schemes where using a gradient-based method is disadvantageous. This lecture gives a gentle introduction to evolutionary techniques, their main operators, possible use cases, and their biological background. Students will get to know how to use mutation, crossover, and selection operators to control exploration and exploitation during a heuristic search. Besides the theory of the basic evolutionary algorithm, state-of-the-art variations (such as different swarm intelligence algorithms) are also introduced.

With Deep Learning being the focus of current AI research we have to ask the question what would be the next step?
A possible answer to this question could be neuromorphic computing and Spiking Neural Networks (SNNs). This session is an introduction to these dynamic, biologically-inspired neural networks, and their properties. We cover the three most common modeling techniques of such neurons namely the Integrate-and-Fire, the Spike Response Model, and the Event-based Model. The basics of Hebbian and gradient-based learning of such neural networks are also introduced. The lecture concludes with a discussion of the latest trends in SNN research and possible applications of ANN-to-SNN conversion.

Presentation aims to promote the standards of research ethics and legal aspects for AI under the frame of the AI PR in order to explore the importance of ethical research and responsible use of AI. The focus relates to research, development, use and impact of AI supported technologies, while we not only discuss the theoretical aspects, but also the sustainable tools available for practical applications. By considering different perspectives, we provide an interdisciplinary attitude for each stakeholder.