-
Deep Learning for Climate Model Output Statistics
Authors:
Michael Steininger,
Daniel Abel,
Katrin Ziegler,
Anna Krause,
Heiko Paeth,
Andreas Hotho
Abstract:
Climate models are an important tool for the assessment of prospective climate change effects but they suffer from systematic and representation errors, especially for precipitation. Model output statistics (MOS) reduce these errors by fitting the model output to observational data with machine learning. In this work, we explore the feasibility and potential of deep learning with convolutional neu…
▽ More
Climate models are an important tool for the assessment of prospective climate change effects but they suffer from systematic and representation errors, especially for precipitation. Model output statistics (MOS) reduce these errors by fitting the model output to observational data with machine learning. In this work, we explore the feasibility and potential of deep learning with convolutional neural networks (CNNs) for MOS. We propose the CNN architecture ConvMOS specifically designed for reducing errors in climate model outputs and apply it to the climate model REMO. Our results show a considerable reduction of errors and mostly improved performance compared to three commonly used MOS approaches.
△ Less
Submitted 9 December, 2020;
originally announced December 2020.
-
Landing together: how flocks arrive at a coherent action in time and space in the presence of perturbations
Authors:
Bence Ferdinandy,
Kunal Bhattacharya,
Daniel Abel,
Tamas Vicsek
Abstract:
Collective motion is abundant in nature, producing a vast amount of phenomena which have been studied in recent years, including the landing of flocks of birds. We investigate the collective decision making scenario where a flock of birds decides the optimal time of landing in the absence of a global leader. We introduce a simple phenomenological model in the spirit of the statistical mechanics-ba…
▽ More
Collective motion is abundant in nature, producing a vast amount of phenomena which have been studied in recent years, including the landing of flocks of birds. We investigate the collective decision making scenario where a flock of birds decides the optimal time of landing in the absence of a global leader. We introduce a simple phenomenological model in the spirit of the statistical mechanics-based self-propelled particles (SPP-s) approach to interpret this process. We expect that our model is applicable to a larger class of spatiotemporal decision making situations than just the landing of flocks (which process is used as a paradigmatic case). In the model birds are only influenced by observable variables, like position and velocity. Heterogeneity is introduced in the flock in terms of a depletion time after which a bird feels increasing bias to move towards the ground. Our model demonstrates a possible mechanism by which animals in a large group can arrive at an egalitarian decision about the time of switching from one activity to another in the absence of a leader. In particular, we show the existence of a paradoxical effect where noise enhances the coherence of the landing process.
△ Less
Submitted 11 March, 2012; v1 submitted 11 February, 2012;
originally announced February 2012.
-
Patterns, transitions and the role of leaders in the collective dynamics of a simple robotic flock
Authors:
Norbert Tarcai,
Csaba Virágh,
Dániel Ábel,
Máté Nagy,
Péter L. Várkonyi,
Gábor Vásárhelyi,
Tamás Vicsek
Abstract:
We have developed an experimental setup of very simple self-propelled robots to observe collective motion emerging as a result of inelastic collisions only. A circular pool and commercial RC boats were the basis of our first setup, where we demonstrated that jamming, clustering, disordered and ordered motion are all present in such a simple experiment and showed that the noise level has a fundamen…
▽ More
We have developed an experimental setup of very simple self-propelled robots to observe collective motion emerging as a result of inelastic collisions only. A circular pool and commercial RC boats were the basis of our first setup, where we demonstrated that jamming, clustering, disordered and ordered motion are all present in such a simple experiment and showed that the noise level has a fundamental role in the generation of collective dynamics. Critical noise ranges and the transition characteristics between the different collective patterns were also examined. In our second experiment we used a real-time tracking system and a few steerable model boats to introduce intelligent leaders into the flock. We demonstrated that even a very small portion of guiding members can determine group direction and enhance ordering through inelastic collisions. We also showed that noise can facilitate and speed up ordering with leaders. Our work was extended with an agent-based simulation model, too, and high similarity between real and simulation results were observed. The simulation results show clear statistical evidence of three states and negative correlation between density and ordered motion due to the onset of jamming. Our experiments confirm the different theoretical studies and simulation results in the literature about collision-based, noise-dependent and leader-driven self-propelled particle systems.
△ Less
Submitted 3 February, 2012;
originally announced February 2012.
-
Modularity measure of networks with overlapping communities
Authors:
Anna Lázár,
Dániel Ábel,
Tamás Vicsek
Abstract:
In this paper we introduce a non-fuzzy measure which has been designed to rank the partitions of a network's nodes into overlapping communities. Such a measure can be useful for both quantifying clusters detected by various methods and during finding the overlapping community-structure by optimization methods. The theoretical problem referring to the separation of overlapping modules is discusse…
▽ More
In this paper we introduce a non-fuzzy measure which has been designed to rank the partitions of a network's nodes into overlapping communities. Such a measure can be useful for both quantifying clusters detected by various methods and during finding the overlapping community-structure by optimization methods. The theoretical problem referring to the separation of overlapping modules is discussed, and an example for possible applications is given as well.
△ Less
Submitted 27 October, 2009;
originally announced October 2009.
-
Centrality properties of directed module members in social networks
Authors:
Peter Pollner,
Gergely Palla,
Daniel Abel,
Andras Vicsek,
Illes J. Farkas,
Imre Derenyi,
Tamas Vicsek
Abstract:
Several recent studies of complex networks have suggested algorithms for locating network communities, also called modules or clusters, which are mostly defined as groups of nodes with dense internal connections. Along with the rapid development of these clustering techniques, the ability of revealing overlaps between communities has become very important as well. An efficient search technique f…
▽ More
Several recent studies of complex networks have suggested algorithms for locating network communities, also called modules or clusters, which are mostly defined as groups of nodes with dense internal connections. Along with the rapid development of these clustering techniques, the ability of revealing overlaps between communities has become very important as well. An efficient search technique for locating overlapping modules is provided by the Clique Percolation Method (CPM) and its extension to directed graphs, the CPMd algorithm. Here we investigate the centrality properties of directed module members in social networks obtained from e-mail exchanges and from sociometric questionnaires. Our results indicate that nodes in the overlaps between modules play a central role in the studied systems. Furthermore, the two different types of networks show interesting differences in the relation between the centrality measures and the role of the nodes in the directed modules.
△ Less
Submitted 3 September, 2008;
originally announced September 2008.
