Key research themes
1. How can computational search algorithms be optimized to reduce surplus node generation and improve memory and time efficiency in range-based pathfinding problems?
This research area investigates improvements over classical A* search algorithms in domains with large branching factors, focusing on minimizing the generation of surplus nodes—nodes with costs exceeding the optimal solution cost—which negatively impact both memory and runtime efficiency. By incorporating heuristic and domain-specific knowledge to control node expansions and avoid unnecessary computations, these improved algorithms target range expansion scenarios effectively, especially where pathfinding is guided by admissible heuristics.
2. What formal methods allow over-approximation and efficient computation of output ranges in neural networks, and how can they facilitate safety verification in range expansions of network outputs?
This theme focuses on formal verification approaches to compute output ranges of feed-forward neural networks, particularly with ReLU activation functions, leveraging abstraction techniques to reduce computational complexity. It addresses how interval neural networks (INNs) can be constructed to over-approximate output ranges, enabling scalability in safety-critical applications where reachable sets and safe operation ranges must be determined under uncertain inputs. The goal is to maintain soundness while improving verification tractability within expanded input and output domains.
3. How do urbanization, land cover, and human population density changes influence large carnivore range expansions in Europe?
This theme investigates correlations between changes in human population density, land cover (forest and agriculture), and legislative protections with the distributional expansions of three large carnivores—grey wolf, Eurasian lynx, and brown bear—across Europe. It integrates multi-temporal species distribution models with landscape and socio-demographic data over two decades to understand habitat suitability changes and how anthropogenic factors promote or restrict large carnivore recolonization and range expansion in fragmented human-dominated territories.
4. What behavioral and ecological mechanisms contribute to successful colonization and range expansion in social species and marine top predators?
This theme explores how animal prospecting behavior, social information use, and habitat selection influence dispersal decisions in social species such as Audouin's gulls, as well as how large-scale breeding range shifts affect foraging ecology and population dynamics in top marine predators like Laysan albatrosses. It synthesizes telemetry and long-term monitoring data to understand how range expansions impact the species' use of novel environments, breeding performance, and foraging strategies in new ecological ranges.
5. How can range reduction and modular arithmetic techniques be utilized to optimize computation of mathematical functions over expanded argument ranges?
This theme encompasses algorithmic strategies for decomposing large-range function evaluations into computations over smaller, manageable argument intervals, employing modular and multiplicative range reduction techniques. These methods enable the usage of polynomial or rational approximations within convergence domains, improving computational accuracy and efficiency for transcendental functions. The focus is on addressing challenges due to argument size, cancellation errors, and efficient reduction steps critical for implementations in numerical and computer arithmetic.