Artificial intelligence
Responsible instructor: Katalin Hangos (DSc), coordinator: Ágnes Vathy-Fogarassy
Exam material: 6 chosen topics from the following list, which the examinee agrees with the examiner in advance
EXAM TOPICS
1. Local searches, trajectory-based algorithms, and population-based algorithms
The aims, operations, and characteristics of various search methods. Local search methods. Characteristics of trajectory-based algorithms (e.g. tabu search, simulated annealing). Population-based algorithms (e.g. PSO, ant colony optimization, evolutionary algorithms).
(Advanced metaheuristic algorithms)
2. Advanced genetic algorithms
Solutions and reasons for applying advanced representations, genetic operators, and selection methods. Adaptive parameter and operator selection. Self-adaptive GA. Hybrid GA. The concept of Pareto front. Pareto dominance, fitness assignment methods: e.g. NSGA-II, SPEA2. Constraint handling with GA: penalty-based approaches, repair mechanisms. Handling complex and conflicting objectives.
(Advanced metaheuristic algorithms)
3. Markov processes and model-based solution methods
Markov Reward Processes and their solutions, Markov Decision Processes and their solutions, comparing the Bellman Expectation Equation and the Bellman Optimality Equation in terms of their role and form, planning by dynamic programming (policy evaluation, policy iteration, and value iteration)
(Reinforcement learning and its applications)
4. Model-free reinforcement learning
Model-free prediction (Monte Carlo Learning, Temporal-Difference Learning, and TD(λ)), Model-free control (on-policy methods: Monte Carlo Learning, on-policy Temporal-Difference Learning, and Sarsa(λ); off-policy Learning: importance sampling and Q-learning)
(Reinforcement learning and its applications)
5. Classification algorithms and their evaluations
Classification algorithms (decision trees, Bayes classification methods, k-nn classification, Linear Discriminant Analysis, binary and multinominal logistic regression, Support Vector Machines). Embedded methods. Evaluation methods and metrics for classification.
(Machine Learning for Predictive Analytics)
6. Regression algorithms and their evaluations
Linear regression, polynomial regression, Ridge regression, Lasso regression, Elastic Net, orthogonal regression, decision tree regressor, k-nearest neighbor regressor. Regression error metrics.
(Machine Learning for Predictive Analytics)
7. Handling typical problems of datasets in machine learning
Handling small and large datasets: oversampling methods (random, SMOTE és variánsai, ADASYN), undersampling methods (random, cluster centroids, Near Miss, Tomek Links, Edited Nearest Neighbor Rule, Neighbourhood Cleaning Rule, Condensed Nearest Neighbor Rule). Combining the oversampling and undersampling. Few-shot learning, one-shot learning, zero-shot learning. Handling datasets with large dimension: Feature selection (filter, wrapper, embedded and hybrid) methods. Measuring feature importance.
(Machine Learning for Predictive Analytics)
8. Neural networks and deep learning in processing structured datasets
Structural design of neural networks, activation functions and their role, the gradient method and other optimization algorithms, loss functions, overfitting, regularization techniques. Recurrent neural networks (RNN, LSTM, GRU, DRNN), the effect of recurrent links.
(Deep Learning and its Applications)
9. The transformer architecture and its variants
Description of BERT, Generative Pre-trained Transformer, XLNet, and Text-to-Text Transfer Transformer architectures and their properties. Detailed explanation of transformer architectures (multi-head self-attention mechanism, positional encoding, their training process). Transfer learning and fine-tuning for applying pre-trained transformers in new domains.
(Generative networks and transformers)
10. Generative Adversarial Networks (GANs) and transformer-based generative models. Embeddings and vector databases.
The structure and training of GANs (the tasks of the generator and discriminator). The structure and operation of transformer-based generative models. The concept of embedding, possible formats of input and their mapping. Specialties and characteristics of vector databases.
(Generative networks and transformers)
11. Convolutional networks and attention mechanisms in image processing.
VGG16, SSD, and Yolo architectures. Attention mechanisms and visual transformers. Contrastive learning.
(Learning algorithms of machine vision)
12. AI techniques used in applications for dynamic systems
The software architecture of intelligent control systems, the properties of the intelligent and real-time subsystems and their cooperation; rule-based expert systems, real-time expert systems, relationship between dynamic and intelligent system models; qualitative differential and difference equations, Petri-nets, fuzzy control systems; analysis of the properties of intelligent dynamic system models.
(Intelligent control systems)
13. Analysis and control of uncertain dynamic systems using AI methods
Dynamic systems with uncertainties and their description using qualitative models; colored Petri net models and their use for constructing and analyzing sceduling procedures; diagnosis using SDG models, fuzzy qualitative simulation and colored Petri nets; control design and analysis using fuzzy rules
(Engineering applications of artificial intelligence)