Iskanje po repozitoriju

Opis: Reinforcement learning has emerged as a prominent method for controlling dynamic systems in the absence of a precise mathematical model. However, its reliance on extensive interactions with the environment often leads to prolonged training periods. In this paper, we propose an alternative approach to learning control policies that focuses on learning qualitative models and uses symbolic planning to derive a qualitative plan for the control task, which is executed by an adaptive reactive controller. We conduct experiments utilizing our approach on the cart-pole problem, a standard benchmark in dynamic system control. We additionally extend this problem domain to include uneven terrains, such as driving over craters or hills, to assess the robustness of learned controllers. Our results indicate that qualitative learning offers significant advantages over reinforcement learning in terms of sample efficiency, transferability, and interpretability. We demonstrate that our proposed approach is at least two orders of magnitude more sample efficient in the cart-pole domain than the usual variants of reinforcement learning.
Ključne besede: qualitative modeling, qualitative reasoning, qualitative control, transfer learning
Objavljeno v RUP: 07.08.2025; Ogledov: 599; Prenosov: 11
Celotno besedilo (1,53 MB)
Gradivo ima več datotek! Več...

Opis: Fisheries bycatch is the main threat to sea turtle populations in marine habitats in the Mediterranean Sea. Bycatch in set nets accounts for 30,000 catches yearly, which combined with mortality rates of up to 82.6 %, makes it the deadliest fishing gear in the region. The Adriatic Sea is one of the most important foraging and developmental habitats for loggerhead sea turtles (Caretta caretta) in the Mediterranean. Bycatch in the Adriatic small-scale set net fishery is estimated at over 8900 individuals per year placing its mitigation among international conservation priorities. Use of light emitting diodes (LED lights) of different wavelengths as bycatch deterrents was proposed as an effective conservation tool, reducing bycatch levels by up to 93 %. In the present study we: (i) quantified the effects of net illumination on loggerhead sea turtle bycatch rates and (ii) demonstrate the importance of regional validation of conservation tools effectiveness prior to their widespread adoption. We assessed the effects of green LED lights on sea turtle bycatch rates in 390 net pairs with total length of 1278 km monitored nets. We found no significant change in sea turtle bycatch rates as a result of net illumination. A significant decrease of 35 % of primary target catch was observed in illuminated trammel nets. Possible explanations include fluctuating turbidity and an adapted LED light attachment method. Observed sea turtle bycatch rates in unilluminated nets were up to 84 times lower than in comparable gears in Northern Adriatic Sea. Understanding the drivers behind the differing bycatch rates would be necessary before conservation decisions are made. Our study showed that the net illumination is not necessarily a silver bullet for bycatch mitigation in set nets globally, and that regional testing is a key requirement before its introduction in fisheries as a conservation tool.
Ključne besede: marine megafauna, Caretta caretta, small scale fisheries, bottom set nets, bycatch hotspot, bycatch mitigation, visual deterrents
Objavljeno v RUP: 06.08.2025; Ogledov: 554; Prenosov: 6
Celotno besedilo (3,58 MB)
Gradivo ima več datotek! Več...

Opis: This opinion paper advocates for project-based learning (PBL) and teaching as a way of preserving active learning in timber engineering education. A recent example of using PBL in a timber engineering course is presented. Its strengths and associated challenges are briefly highlighted, and some suggestions are provided for the adoption of such approaches.
Ključne besede: project-based learning, timber engineering, engineering higher education
Objavljeno v RUP: 06.08.2025; Ogledov: 567; Prenosov: 2
Celotno besedilo (452,64 KB)

Opis: We continue the study of balanceable graphs, defined by Caro, Hansberg, and Montejano in 2021 as graphs G such that any 2-coloring of the edges of a sufficiently large complete graph containing sufficiently many edges of each color contains a balanced copy of G (that is, a copy with half the edges of each color). While the problem of recognizing balanceable graphs was conjectured to be NP-complete by Dailly, Hansberg, and Ventura in 2021, balanceable graphs admit an elegant combinatorial characterization: a graph is balanceable if and only there exist two vertex subsets, one containing half of all the graph’s edges and another one such that the corresponding cut contains half of all the graph’s edges. We consider a special case of this property, namely when one of the two sets is a vertex cover, and call the corresponding graphs simply balanceable. We prove a number of results on balanceable and simply balanceable regular graphs. First, we characterize simply balanceable regular graphs via a condition involving the independence number of the graph. Second, we address a question of Dailly, Hansberg, and Ventura from 2021 and show that every cubic graph is balanceable. Third, using Brooks’ theorem, we show that every 4-regular graph with order divisible by 4 is balanceable. Finally, we show that it is NP-complete to determine if a 9-regular graph is simply balanceable.
Ključne besede: balanceable graph, simply balanceable graph, cubic graph, 4-regular graph, regular graph, independent set
Objavljeno v RUP: 06.08.2025; Ogledov: 576; Prenosov: 7
Celotno besedilo (530,38 KB)
Gradivo ima več datotek! Več...

Opis: Background Vector-borne diseases are becoming increasingly important in public health. The incidence of Lyme borreliosis in Slovenia continues to rise. Therefore, screening the prevalence of Borrelia spp. in vectors and their hosts is very important. Rarely, but still, it is possible to discover new species or subtypes in a new region. Methods As part of a master’s thesis, a study group collected ticks in a coastal area of Slovenia. The ticks were incubated in an MKP medium for a possible Borrelia culture. Borrelia DNA was extracted from positive cultures. We performed classical molecular genotyping methods to complement whole genome sequencing (WGS) analysis. Results Using classical genotyping, we confirmed previously undetected Borrelia species in a Slovenian tick, Borrelia lusitanie. Using large restriction fragment polymorphisms (LRFP) with the enzyme MluI, we were able to detect a new LRFP subtype of Borrelia lusitanie, namely Mll3. Using WGS, we identified a new MLST type (ST-1154) and reconstructed five linear plasmids (lp17, lp25, lp28–8, lp54 and lp56+38) and two circular plasmids (cp26 and cp32–11+1). Conclusion Regular screening of vectors and their hosts for the presence of Borrelia pathogens is important to detect the spread of different species in new areas. When reconstructing Borrelia plasmids, it is very important to apply different assembly methods and to critically analyse the results in order to select suitable contigs for further trimming and to successfully generate a complete consensus genome of high quality from hybrid read data.
Ključne besede: Slovenia, Borrelia lusitaniae, tick, Ixodes ricinus, WGS, plasmid analysis
Objavljeno v RUP: 05.08.2025; Ogledov: 513; Prenosov: 2
Celotno besedilo (580,15 KB)
Gradivo ima več datotek! Več...