|
|
Reactions of avian predators to automimicry in milkweed bugs (Heteroptera: Lygaeidae):
Stránská, Anna ; Exnerová, Alice (advisor) ; Baňař, Petr (referee)
Automimicry, or intraspecific variation in defence mechanisms in aposematic prey, is very common in nature. Especially in chemically protected prey. This study contains two experiments. The first experiment investigated the reaction of naive predators, which were Great tits (Parus major), to automimetic prey. The prey was the black-and-red-bug (Lygaeus equestris). The tits were divided into three experimental groups and each group was presented with three bugs that were fed on an artificial diet with different amounts of cardenolides. The control group was presented with prey that had no cardenolides in the diet. The group tested with the low concentration received bugs that fed on a diet with a low concentration of cardenolides and the group tested with the high concentration received bugs that fed on a diet with a high concentration of cardenolides. In a generalization test, all groups were then offered a single firebug (Pyrrhocoris apterus). It was found that the group tested with the high concentration experienced a higher rate of aversive learning. This group also killed and consumed fewer firebugs than the other two groups. The group tested with high concentration generalized most to novel red-and-black prey because they were least likely to attack the firebug. The second experiment tested...
|
|
|
Reactions of predators towards species of red-and-black mimetic complex
Kotlíková, Lucie ; Exnerová, Alice (advisor) ; Baňař, Petr (referee)
The red-black mimetic complex in the Western Palaearctic region includes a large number of arthropod species. These species differ in the degree of their mimetic resemblance, as well as in defensive mechanisms and their effectiveness against various predators. This study is based on two experiments. The first experiment was carried out with adult great tits (Parus major) and artificial prey (photographs). The birds were divided into two experimental groups and were trained to discriminate between palatable and unpalatable prey. One group was trained with higher diversity in the coloration of unpalatable prey (ten species of subfamily Lygaeinae), while the other was trained with low diversity in the coloration of palatable prey (ten individuals of the same species, Lygaeus equestris). After ten learning blocks, two generalization blocks followed, in which both groups received the same prey that was completely different from the prey during learning phase. The rate of learning was not significantly different between the two groups. However, more effective generalization was observed in the group trained with higher prey diversity. However, this trend was only observed in the first generalization block. On the second day, both groups achieved similar generalization success. The second experiment was...
|
|
|
Výstražná signalizace barevných forem slunéčka východního (\kur{Harmonia axyridis})
BOROVIČKA, Martin
The efficiency of the aposematic defenses (colouration and chemical defence) used by three rare colour forms of Harmonia axyridis were tested using in the wild caught great tits (Parus major) as predators. I predicted that visually orienting predator will pay more attention to the rare colour forms, as they are not familiar with them and need to generalize the learned aversion to some other ladybirds to these novel ones.
|
|
|
Artificial Neural Networks and Their Usage For Knowledge Extraction
Petříčková, Zuzana ; Mrázová, Iveta (advisor) ; Procházka, Aleš (referee) ; Andrejková, Gabriela (referee)
Title: Artificial Neural Networks and Their Usage For Knowledge Extraction Author: RNDr. Zuzana Petříčková Department: Department of Theoretical Computer Science and Mathema- tical Logic Supervisor: doc. RNDr. Iveta Mrázová, CSc., Department of Theoretical Computer Science and Mathematical Logic Abstract: The model of multi/layered feed/forward neural networks is well known for its ability to generalize well and to find complex non/linear dependencies in the data. On the other hand, it tends to create complex internal structures, especially for large data sets. Efficient solutions to demanding tasks currently dealt with require fast training, adequate generalization and a transparent and simple network structure. In this thesis, we propose a general framework for training of BP/networks. It is based on the fast and robust scaled conjugate gradient technique. This classical training algorithm is enhanced with analytical or approximative sensitivity inhibition during training and enforcement of a transparent in- ternal knowledge representation. Redundant hidden and input neurons are pruned based on internal representation and sensitivity analysis. The performance of the developed framework has been tested on various types of data with promising results. The framework provides a fast training algorithm,...
|
|
|
Commented translation: Guide de manoeuvre (Éric Tabarly, Luçon: Éditions Le Télégramme, 2008)
Kofroňová, Zuzana ; Belisová, Šárka (advisor) ; Šotolová, Jovanka (referee)
The thesis consists of two main parts- translation and its commentary. It is a translation of 3 chapters of a French yachtsmen guide Guide de manoeuvre (Wharf mooring or double moor to another boat, Anchoring and Man overboard maneuver). In my translation, I tried to preserve the major function of the original text - to inform and instruct the reader. The second part of the work is a commentary of the translation which is focused on translation analysis of the original text, typology of translation problems and the methods of the translation. In the translation analysis I describe the original text with the factors affecting the text. Then I focus on the translation problems I dealt with and how they were solved. In the conclusion, I describe the translations methods I used.
|
|
| |
|
|
Effect of chromatic component on function of antipredatory warning signals
Truhlářová, Marie ; Exnerová, Alice (advisor) ; Pipek, Pavel (referee)
Warning coloration is used by prey to signal its unprofitability to potential predators. Warning colours may have different effects on various cognitive processes of predators (innate avoidance, avoidance learning, memory and generalization). Typical colours regarded as aposematic are red, orange and yellow. Red colour is considered to be the most effective signal and has a significant effect on avoidance learning, memory and generalization. Orange also represents an effective warning signal, though it has been studied less. Yellow is effective aposematic stimulus but it has frequently been found less effective compared to red and orange. Warning functions of white, blue, violet and ultraviolet colours were studied less frequently and their role in aposematism is not yet clear. Iridescent coloration might also be an effective warning signal affecting avoidance learning, memory and generalization. In this thesis I present a summary of information with regard to different warning colours and their effect on cognitive processes of predators. Key words: aposematism, warning coloration, cognitive processes, red, orange, yellow, iridescence, avoidance learning, innate avoidance, memory, generalization
|
|
|
Smoothness of Functions Learned by Neural Networks
Volhejn, Václav ; Musil, Tomáš (advisor) ; Straka, Milan (referee)
Modern neural networks can easily fit their training set perfectly. Surprisingly, they generalize well despite being "overfit" in this way, defying the bias-variance trade-off. A prevalent explanation is that stochastic gradient descent has an implicit bias which leads it to learn functions that are simple, and these simple functions generalize well. However, the specifics of this implicit bias are not well understood. In this work, we explore the hypothesis that SGD is implicitly biased towards learning functions that are smooth. We propose several measures to formalize the intuitive notion of smoothness, and conduct experiments to determine whether these measures are implicitly being optimized for. We exclude the possibility that smoothness measures based on first derivatives (the gradient) are being implicitly optimized for. Measures based on second derivatives (the Hessian), on the other hand, show promising results. 1
|
|
|
Generalization of LOD2 building models using the aggregation method
Měchurová, Kristýna ; Brůha, Lukáš (advisor) ; Pokorný, Tomáš (referee)
Generalization of LOD2 building models using the aggregation method Abstract The thesis proposes and implements a method of 3D building models aggregation. The procedure achieves global optima by the means of mathematic optimization. Buildings are aggregated according to similarity characteristics typical for LOD2, e. g. roof type. Aggregation process is driven by minimalization of volume changes and of the aggregate count. The optimization problem was implemented as a Python script with optional parameters to meet custom demands of a wide range of users. Input data models of buildings are created by the method of procedural modelling. Its outcome is further restructured into form of continuous blocks. Finally, the visualization procedure is designed and implemented to illustrate the results of optimized aggregation of 3D building models. Keywords: 3D GIS, generalization, aggregation, mathematical optimization, procedural modelling
|
|
|
Simplification of buildings based on typification
Gottstein, Otomar ; Bayer, Tomáš (advisor) ; Jindrák, Přemysl (referee)
Simplification of buildings based on typification Abstract The diploma thesis is focused on cartographic generalization. Its main aim was to develop a new method of simplification of buildings based on typification for areas with lower density of buildings (rural or mountain areas). The proposed method was designed for large scale maps (1 : 25 000 and 1 : 50 000). The presented generalization algorithm is based on preferential selection of buildings to be drawn on the map according to their civic importance, area and location towards roads, railways and streams. It respects cartographic rules used for this type of generalization. The algorithm was implemented in the Python programming language using the Shapely and Fiona libraries for a purpose of its proper testing. The ZABAGED and DATA50 data were chosen as suitable test data. Among other things, this thesis also introduces the evaluation method of typification results, which uses Voronoi diagram. Achieved results are presented on maps of fifteen villages with different spatial structure in both targeted scales. Keywords: digital cartography, generalization, typification, building, simplification
|
guest ::
