Abstract: By action model, we understand any logic-based representation of effects and executability preconditions of individual actions within a certain domain. In the context of artificial intelligence, such models are necessary for planning and goal-oriented automated behaviour. Currently, action models are commonly hand-written by domain experts in advance. However, since this process is often difficult, time-consuming, and error-prone, it makes sense to let agents learn the effects and conditions of actions from their own observations. Even though the research in the area of action learning, as a certain kind of inductive reasoning, is relatively young, there already exist several distinctive action learning methods. We will try to identify the collection of the most important properties of these methods, or challenges that they are trying to overcome, and briefly outline their impact on practical applications., Abstrakt: Podle akčního modelu chápeme logickou reprezentaci efektů a předpokladů vykonatelnosti jednotlivých akcí v rámci určité domény. V kontextu umělé inteligence jsou tyto modely nezbytné pro plánování a cílené automatizované chování. V současné době jsou akční modely běžně ručně psány odborníky domény předem. Vzhledem k tomu, že tento proces je často obtížný, časově náročný a náchylný k chybám, má smysl nechat agenty seznámit se s účinky a podmínkami akcí z vlastních pozorování. I když je výzkum v oblasti akčního učení, jako určitý druh indukčního uvažování, relativně mladý, existuje již několik výrazných metod učení. Pokusíme se identifikovat sbírku nejdůležitějších vlastností těchto metod., and Michal Čertický
The paper deals with a description of a constructive neural network based on gradient initial setting of its weights. The network has been used as a pattern classifier of two dimensional patterns but it can be generally used to n x m associative problems. A network topology, processes of learning and retrieving, experiments and comparison to other neural networks are described.
Extremely low-frequency magnetic field (ELF-MF) has been suggested to influence the cognitive capability but this should be dynamically evaluated in a longitudinal study. Previous training can affect performance, but the influence under magnetic field is unclear. This study aims to evaluate the effects of previous training and ELF-MF exposure on learning and memory using the Morris water maze (MWM). Sprague-Dawley rats were subjected to MWM training, ELF-MF exposure (50 Hz, 100 μT), or ELF-MF exposure combined with MWM training for 90 days. Normal rats were used as controls. The MWM was used to test. The data show that the rats exposed to training and ELF-MF with training performed better on spatial acquisition when re-tested. However, during the probe trial the rats showed no change between the training phase and the test phase. Compared with the control group, the ELF-MF group showed no significant differences. These results confirm that previous training can improve the learning and memory capabilities regarding spatial acquisition in the MWM and this effect can last for at least 90 days. However, this improvement in learning and memory capabilities was not observed during the probe trial. Furthermore, ELF-MF exposure did not interfere with the improvement in learning and memory capabilities., Y. Li, C. Zhang, T. Song., and Obsahuje bibliografii
We show that the learning of (uncertain) conditional and/or causal information may be modelled by (Jeffrey) imaging on Stalnaker conditionals. We adapt the method of learning uncertain conditional information proposed in Günther (2017) to a method of learning uncertain causal information. The idea behind the adaptation parallels Lewis (1973c)’s analysis of causal dependence. The combination of the methods provides a unified account of learning conditional and causal information that manages to clearly distinguish between conditional, causal and conjunctive information. Moreover, our framework seems to be the first general solution that generates the correct predictions for Douven (2012)’s benchmark examples and the Judy Benjamin Problem., Ukazujeme, že učení (neurčitých) podmíněných a / nebo kauzálních informací může být modelováno zobrazením (Jeffrey) na Stalnakerových podmínkách. Metodu učení nejistých podmíněných informací navrhovaných v Güntheru (2017) přizpůsobujeme metodě učení nejistých kauzálních informací. Myšlenka adaptačních paralel Lewisova (1973c) analýza kauzální závislosti. Kombinace metod poskytuje jednotný popis učení podmíněných a příčinných informací, které dokáží jasně rozlišit mezi podmíněnými, příčinnými a spojovacími informacemi. Náš rámec se navíc jeví jako první obecné řešení, které vytváří správné předpovědi pro příklady benchmarku Douven (2012) a problém Judy Benjaminové., and Mario Günther
During the early postnatal age environmental signals underlie the development of sensory systems. The visual system is considered as an appropriate system to evaluate role of sensory experience in postnatal development of sensory systems. This study was made to assess the effect of visual deprivation on strategy of arm selection in navigation of radial arm maze. Six-week-old light- (LR, control) and dark-reared (DR) rats were trained for correct choices and adjacent arms tasks. Our results showed that both the LR and DR animals equally selected correct arms. In the adjacent arms task, however, the control group significantly outperformed the DR animals. While the LR males and females displayed some differences in performing the tasks, no sex dependency was found in the performance of the DR group. These findings indicate that the lack of visual experience is likely to influence the strategy selection as well as sex differences. Thus the difference in the performance of LR and DR animals seems to be due to the male rather than female behavior., M. Salami., and Obsahuje bibliografii a bibliografické odkazy