Real problem solving by artificial intelligence techniques – Faculty of Mathematics and Computer Science

Introduction to Artificial Intelligence, professor Horia F. Pop, PhD
The course focuses on defining the field of Artificial Intelligence by association with human intelligence, with reference to the problem-solving qualities of humans. We will also look at the difference between AI-based solutions and standard solutions. We will then illustrate a set of introductory algorithms in the field of Artificial Intelligence, namely search algorithms in a statespace.
Optimisation techniques, professor Camelia Chira, PhD
Optimisation challenges come up in many actual applications in industry and society. Optimisation involves selecting the best decision or solution from a crowd of possible solutions (which is usually too wide to be able to test all possibilities and then choose the optimal one). Complex problems such as finding the shortest route in traffic, planning tasks, allocating resources or optimising a financial portfolio require the use of intelligent optimisation techniques. The course introduces some Artificial Intelligence techniques that can be useful in optimisation and exemplifies their use in solving complex real-world problems.
Supervised Machine Learning, professor Laura Dioșan, PhD
Society today is confronted with problems such as determining the price of a product, tailoring a marketing strategy to a customer’s profile, traffic sign recognition or estimating the risk of cervical cancer in women.
While each problem comes with its own challenges, artificial intelligence and supervised learning techniques can identify feasible solutions to these problems. The solutions built interactively in the course will be analysed from both extrinsic (customer/business perspective) and intrinsic (data analyst view) perspectives.
Unsupervised Machine Learning, teaching assistant Bogdan Mursa, doctoral student
Most of the process above usually require datasets specifically annotated by specialists in the target domains (marketing experts, doctors, etc.), which leads to higher costs and sometimes even to the impossibility of applying supervised learning. An alternative in these cases is unsupervised learning, which can work with more diverse data that has not been specifically developed to be used by certain algorithms. The course will present, interactively, the main algorithms in unsupervised learning, focusing on understanding them conceptually.