Nordic Machine Intelligence https://journals.uio.no/NMI <p>Nordic Machine Intelligence (NMI) is a non-commercial, open-access , peer-reviewed journal and thus qualifies to be a diamond open-access journal. The journal publishes original research articles, literature reviews, conference articles related to NORA's Norwegian and Nordic conferences, articles related to the <a href="https://www.nora.ai/Competition/">NMI Challenge</a>, statements and other educational material within all aspects of artificial intelligence.</p> en-US anne.hakansson@uit.no (Anne Håkansson) b.j.singstad@fys.uio.no (Bjørn-jostein Singstad) Fri, 03 Jun 2022 00:00:00 +0200 OJS 3.3.0.11 http://blogs.law.harvard.edu/tech/rss 60 Hierarchical Object Detection applied to Fish Species https://journals.uio.no/NMI/article/view/9452 <p>Gathering information of aquatic life is often based on timeconsuming<br />methods utilizing video feeds. It would be beneficial<br />to capture more information cost-effectively from video feeds.<br />Video based object detection has an ability to achieve this.<br />Recent research has shown promising results with the use of<br />YOLO for object detection of fish. As underwater conditions<br />can be difficult and thus fish species are hard to discriminate.<br />This study proposes a hierarchical structure-based YOLO Fish<br />algorithm in both the classification and the dataset to gain<br />valuable information. With the use of hierarchical classification<br />and other techniques. YOLO Fish is a state-of-the-art object<br />detector on Nordic fish species, with an mAP of 91.8%. The<br />algorithm has an inference time of 26.4 ms, fast enough to<br />run on real-time video on the high-end GPU Tesla V100.</p> Dr. Aditya Gupta, Espen Stausland Kalhagen, Ørjan Langøy Olsen, Dr. Morten Goodwin Copyright (c) 2022 Nordic Machine Intelligence https://creativecommons.org/licenses/by/4.0 https://journals.uio.no/NMI/article/view/9452 Fri, 03 Jun 2022 00:00:00 +0200 A scalable, black-box hybrid genetic algorithm for continuous multimodal optimization in moderate dimensions https://journals.uio.no/NMI/article/view/9633 <p>Optimization problems can be found in many areas of science<br>and technology. Not only the global optimum, but also a<br>(large) number of near-optima are often of interest. This<br>gives rise to what are referred to as multimodal optimization<br>problems. In most cases, the number and quality of the optima<br>are unknown and assumptions cannot be made about the<br>objective functions. In this paper, we focus on continuous,<br>unconstrained optimization in moderately high-dimensional<br>continuous spaces (d ≤ 10).<br>We present a scalable<br>algorithm with virtually no parameters, which performs well<br>for general objective functions (non-convex, discontinuous).<br>It is based on two well-established algorithms (CMA-ES,<br>deterministic crowding). Novel elements of the algorithm<br>include the detection of seed points for local searches and<br>collision avoidance, both based on nearest neighbors, and a<br>strategy for semi-sequential optimization to realize scalability.<br>The performance of the proposed algorithm is numerically<br>evaluated using the CEC2013 niching benchmark suite for<br>1 − 20 dimensional functions, and a 9 dimensional real-world<br>problem from constraint optimization in climate research.<br>The algorithm performs well in relation to the CEC2013<br>benchmarks and only falls short on higher dimensional and<br>strongly inisotropic problems. In the case of the climate-<br>related problem, the algorithm is able to find a high number<br>(&gt; 150) of optima of relevance to climate research. The<br>proposed algorithm does not require special configuration for<br>the optimization problems considered in this paper, i.e., it<br>shows good black-box behavior.</p> Klaus Johannsen, Nadine Goris, Jerry Tjiputra, Bjørnar Jensen Copyright (c) 2022 Nordic Machine Intelligence https://creativecommons.org/licenses/by/4.0 https://journals.uio.no/NMI/article/view/9633 Sat, 26 Nov 2022 00:00:00 +0100 NeuroAI - A strategic opportunity for Norway and Europe https://journals.uio.no/NMI/article/view/9950 <p>At an energy consumption of merely 20 watts, the brain is able to learn, perceive its surroundings, analyze solutions, predict the future and choose actions in a much more general way than any computer algorithm and infrastructure. NeuroAI, an emerging field at the intersection of brain sciences and artificial intelligence, is expected to become a large research field in the years to come. The main promise of NeuroAI is that by understanding how the brain and biological neural networks compute, it will be possible to identify the key components of human intelligence in order to drive the development of a more energy-efficient and flexible artificial intelligence (AI) that will match, perhaps even surpass, human intelligence. NeuroAI is a truly multidisciplinary effort, spanning across several disciplines such as computer science, neuroscience, psychology, philosophy, linguistics, law and ethics. Norway has unique world-class expertise in brain sciences, and a fast growing, well-organized AI community at the highest international level. In this opinion article, we argue that Norway has competitive advantages within NeuroAI, and that Norway has the ideal ecosystem to take a leading role in NeuroAI initiatives in Europe. Norwegian investments in NeuroAI may be a strategic initiative to position Norway in the forefront of AI research worldwide.</p> Mikkel Lepperød, Klas Pettersen, Solve Sæbø, Stefano Nichele Copyright (c) 2022 Nordic Machine Intelligence https://creativecommons.org/licenses/by/4.0 https://journals.uio.no/NMI/article/view/9950 Mon, 28 Nov 2022 00:00:00 +0100