@phdthesis{Elsaid2022,
  author    = {Elsaid, Mohamed Esameldin Mohamed},
  title     = {Virtual machines live migration cost modeling and prediction},
  doi       = {10.25932/publishup-54001},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-540013},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xiv, 107},
  year      = {2022},
  abstract  = {Dynamic resource management is an essential requirement for private and public cloud computing environments. With dynamic resource management, the physical resources assignment to the cloud virtual resources depends on the actual need of the applications or the running services, which enhances the cloud physical resources utilization and reduces the offered services cost. In addition, the virtual resources can be moved across different physical resources in the cloud environment without an obvious impact on the running applications or services production. This means that the availability of the running services and applications in the cloud is independent on the hardware resources including the servers, switches and storage failures. This increases the reliability of using cloud services compared to the classical data-centers environments. In this thesis we briefly discuss the dynamic resource management topic and then deeply focus on live migration as the definition of the compute resource dynamic management. Live migration is a commonly used and an essential feature in cloud and virtual data-centers environments. Cloud computing load balance, power saving and fault tolerance features are all dependent on live migration to optimize the virtual and physical resources usage. As we will discuss in this thesis, live migration shows many benefits to cloud and virtual data-centers environments, however the cost of live migration can not be ignored. Live migration cost includes the migration time, downtime, network overhead, power consumption increases and CPU overhead. IT admins run virtual machines live migrations without an idea about the migration cost. So, resources bottlenecks, higher migration cost and migration failures might happen. The first problem that we discuss in this thesis is how to model the cost of the virtual machines live migration. Secondly, we investigate how to make use of machine learning techniques to help the cloud admins getting an estimation of this cost before initiating the migration for one of multiple virtual machines. Also, we discuss the optimal timing for a specific virtual machine before live migration to another server. Finally, we propose practical solutions that can be used by the cloud admins to be integrated with the cloud administration portals to answer the raised research questions above. Our research methodology to achieve the project objectives is to propose empirical models based on using VMware test-beds with different benchmarks tools. Then we make use of the machine learning techniques to propose a prediction approach for virtual machines live migration cost. Timing optimization for live migration is also proposed in this thesis based on using the cost prediction and data-centers network utilization prediction. Live migration with persistent memory clusters is also discussed at the end of the thesis. The cost prediction and timing optimization techniques proposed in this thesis could be practically integrated with VMware vSphere cluster portal such that the IT admins can now use the cost prediction feature and timing optimization option before proceeding with a virtual machine live migration. Testing results show that our proposed approach for VMs live migration cost prediction shows acceptable results with less than 20\% prediction error and can be easily implemented and integrated with VMware vSphere as an example of a commonly used resource management portal for virtual data-centers and private cloud environments. The results show that using our proposed VMs migration timing optimization technique also could save up to 51\% of migration time of the VMs migration time for memory intensive workloads and up to 27\% of the migration time for network intensive workloads. This timing optimization technique can be useful for network admins to save migration time with utilizing higher network rate and higher probability of success. At the end of this thesis, we discuss the persistent memory technology as a new trend in servers memory technology. Persistent memory modes of operation and configurations are discussed in detail to explain how live migration works between servers with different memory configuration set up. Then, we build a VMware cluster with persistent memory inside server and also with DRAM only servers to show the live migration cost difference between the VMs with DRAM only versus the VMs with persistent memory inside.},
  language  = {en}
}
@phdthesis{Stone2020,
  author    = {Stone, Kate},
  title     = {Predicting long-distance lexical content in German verb-particle constructions},
  doi       = {10.25932/publishup-47679},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-476798},
  school      = {Universit{\"a}t Potsdam},
  year      = {2020},
  abstract  = {A large body of research now supports the presence of both syntactic and lexical predictions in sentence processing. Lexical predictions, in particular, are considered to indicate a deep level of predictive processing that extends past the structural features of a necessary word (e.g. noun), right down to the phonological features of the lexical identity of a specific word (e.g. /kite/; DeLong et al., 2005). However, evidence for lexical predictions typically focuses on predictions in very local environments, such as the adjacent word or words (DeLong et al., 2005; Van Berkum et al., 2005; Wicha et al., 2004). Predictions in such local environments may be indistinguishable from lexical priming, which is transient and uncontrolled, and as such may prime lexical items that are not compatible with the context (e.g. Kukona et al., 2014). Predictive processing has been argued to be a controlled process, with top-down information guiding preactivation of plausible upcoming lexical items (Kuperberg \& Jaeger, 2016). One way to distinguish lexical priming from prediction is to demonstrate that preactivated lexical content can be maintained over longer distances. In this dissertation, separable German particle verbs are used to demonstrate that preactivation of lexical items can be maintained over multi-word distances. A self-paced reading time and an eye tracking experiment provide some support for the idea that particle preactivation triggered by a verb and its context can be observed by holding the sentence context constant and manipulating the predictabilty of the particle. Although evidence of an effect of particle predictability was only seen in eye tracking, this is consistent with previous evidence suggesting that predictive processing facilitates only some eye tracking measures to which the self-paced reading modality may not be sensitive (Staub, 2015; Rayner1998). Interestingly, manipulating the distance between the verb and the particle did not affect reading times, suggesting that the surprisal-predicted faster reading times at long distance may only occur when the additional distance is created by information that adds information about the lexical identity of a distant element (Levy, 2008; Grodner \& Gibson, 2005). Furthermore, the results provide support for models proposing that temporal decay is not major influence on word processing (Lewandowsky et al., 2009; Vasishth et al., 2019). In the third and fourth experiments, event-related potentials were used as a method for detecting specific lexical predictions. In the initial ERP experiment, we found some support for the presence of lexical predictions when the sentence context constrained the number of plausible particles to a single particle. This was suggested by a frontal post-N400 positivity (PNP) that was elicited when a lexical prediction had been violated, but not to violations when more than one particle had been plausible. The results of this study were highly consistent with previous research suggesting that the PNP might be a much sought-after ERP marker of prediction failure (DeLong et al., 2011; DeLong et al., 2014; Van Petten \& Luka, 2012; Thornhill \& Van Petten, 2012; Kuperberg et al., 2019). However, a second experiment in a larger sample experiment failed to replicate the effect, but did suggest the relationship of the PNP to predictive processing may not yet be fully understood. Evidence for long-distance lexical predictions was inconclusive. The conclusion drawn from the four experiments is that preactivation of the lexical entries of plausible upcoming particles did occur and was maintained over long distances. The facilitatory effect of this preactivation at the particle site therefore did not appear to be the result of transient lexical priming. However, the question of whether this preactivation can also lead to lexical predictions of a specific particle remains unanswered. Of particular interest to future research on predictive processing is further characterisation of the PNP. Implications for models of sentence processing may be the inclusion of long-distance lexical predictions, or the possibility that preactivation of lexical material can facilitate reading times and ERP amplitude without commitment to a specific lexical item.},
  language  = {en}
}
@phdthesis{Schlenter2019,
  author    = {Schlenter, Judith},
  title     = {Predictive language processing in late bilinguals},
  doi       = {10.25932/publishup-43249},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-432498},
  school      = {Universit{\"a}t Potsdam},
  pages     = {251},
  year      = {2019},
  abstract  = {The current thesis examined how second language (L2) speakers of German predict upcoming input during language processing. Early research has shown that the predictive abilities of L2 speakers relative to L1 speakers are limited, resulting in the proposal of the Reduced Ability to Generate Expectations (RAGE) hypothesis. Considering that prediction is assumed to facilitate language processing in L1 speakers and probably plays a role in language learning, the assumption that L1/L2 differences can be explained in terms of different processing mechanisms is a particularly interesting approach. However, results from more recent studies on the predictive processing abilities of L2 speakers have indicated that the claim of the RAGE hypothesis is too broad and that prediction in L2 speakers could be selectively limited. In the current thesis, the RAGE hypothesis was systematically put to the test. In this thesis, German L1 and highly proficient late L2 learners of German with Russian as L1 were tested on their predictive use of one or more information sources that exist as cues to sentence interpretation in both languages, to test for selective limits. The results showed that, in line with previous findings, L2 speakers can use the lexical-semantics of verbs to predict the upcoming noun. Here the level of prediction was more systematically controlled for than in previous studies by using verbs that restrict the selection of upcoming nouns to the semantic category animate or inanimate. Hence, prediction in L2 processing is possible. At the same time, this experiment showed that the L2 group was slower/less certain than the L1 group. Unlike previous studies, the experiment on case marking demonstrated that L2 speakers can use this morphosyntactic cue for prediction. Here, the use of case marking was tested by manipulating the word order (Dat > Acc vs. Acc > Dat) in double object constructions after a ditransitive verb. Both the L1 and the L2 group showed a difference between the two word order conditions that emerged within the critical time window for an anticipatory effect, indicating their sensitivity towards case. However, the results for the post-critical time window pointed to a higher uncertainty in the L2 group, who needed more time to integrate incoming information and were more affected by the word order variation than the L1 group, indicating that they relied more on surface-level information. A different cue weighting was also found in the experiment testing whether participants predict upcoming reference based on implicit causality information. Here, an additional child L1 group was tested, who had a lower memory capacity than the adult L2 group, as confirmed by a digit span task conducted with both learner groups. Whereas the children were only slightly delayed compared to the adult L1 group and showed the same effect of condition, the L2 speakers showed an over-reliance on surface-level information (first-mention/subjecthood). Hence, the pattern observed resulted more likely from L1/L2 differences than from resource deficits. The reviewed studies and the experiments conducted show that L2 prediction is affected by a range of factors. While some of the factors can be attributed to more individual differences (e.g., language similarity, slower processing) and can be interpreted by L2 processing accounts assuming that L1 and L2 processing are basically the same, certain limits are better explained by accounts that assume more substantial L1/L2 differences. Crucially, the experimental results demonstrate that the RAGE hypothesis should be refined: Although prediction as a fast-operating mechanism is likely to be affected in L2 speakers, there is no indication that prediction is the dominant source of L1/L2 differences. The results rather demonstrate that L2 speakers show a different weighting of cues and rely more on semantic and surface-level information to predict as well as to integrate incoming information.},
  language  = {en}
}
@phdthesis{Quade2018,
  author    = {Quade, Markus},
  title     = {Symbolic regression for identification, prediction, and control of dynamical systems},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-419790},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xiii, 134},
  year      = {2018},
  abstract  = {In the present work, we use symbolic regression for automated modeling of dynamical systems. Symbolic regression is a powerful and general method suitable for data-driven identification of mathematical expressions. In particular, the structure and parameters of those expressions are identified simultaneously. We consider two main variants of symbolic regression: sparse regression-based and genetic programming-based symbolic regression. Both are applied to identification, prediction and control of dynamical systems. We introduce a new methodology for the data-driven identification of nonlinear dynamics for systems undergoing abrupt changes. Building on a sparse regression algorithm derived earlier, the model after the change is defined as a minimum update with respect to a reference model of the system identified prior to the change. The technique is successfully exemplified on the chaotic Lorenz system and the van der Pol oscillator. Issues such as computational complexity, robustness against noise and requirements with respect to data volume are investigated. We show how symbolic regression can be used for time series prediction. Again, issues such as robustness against noise and convergence rate are investigated us- ing the harmonic oscillator as a toy problem. In combination with embedding, we demonstrate the prediction of a propagating front in coupled FitzHugh-Nagumo oscillators. Additionally, we show how we can enhance numerical weather predictions to commercially forecast power production of green energy power plants. We employ symbolic regression for synchronization control in coupled van der Pol oscillators. Different coupling topologies are investigated. We address issues such as plausibility and stability of the control laws found. The toolkit has been made open source and is used in turbulence control applications. Genetic programming based symbolic regression is very versatile and can be adapted to many optimization problems. The heuristic-based algorithm allows for cost efficient optimization of complex tasks. We emphasize the ability of symbolic regression to yield white-box models. In contrast to black-box models, such models are accessible and interpretable which allows the usage of established tool chains.},
  language  = {en}
}
@phdthesis{Jaeger2015,
  author    = {J{\"a}ger, Lena Ann},
  title     = {Working memory and prediction in human sentence parsing},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-82517},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xi, 144},
  year      = {2015},
  abstract  = {This dissertation investigates the working memory mechanism subserving human sentence processing and its relative contribution to processing difficulty as compared to syntactic prediction. Within the last decades, evidence for a content-addressable memory system underlying human cognition in general has accumulated (e.g., Anderson et al., 2004). In sentence processing research, it has been proposed that this general content-addressable architecture is also used for language processing (e.g., McElree, 2000). Although there is a growing body of evidence from various kinds of linguistic dependencies that is consistent with a general content-addressable memory subserving sentence processing (e.g., McElree et al., 2003; VanDyke2006), the case of reflexive-antecedent dependencies has challenged this view. It has been proposed that in the processing of reflexive-antecedent dependencies, a syntactic-structure based memory access is used rather than cue-based retrieval within a content-addressable framework (e.g., Sturt, 2003). Two eye-tracking experiments on Chinese reflexives were designed to tease apart accounts assuming a syntactic-structure based memory access mechanism from cue-based retrieval (implemented in ACT-R as proposed by Lewis and Vasishth (2005). In both experiments, interference effects were observed from noun phrases which syntactically do not qualify as the reflexive's antecedent but match the animacy requirement the reflexive imposes on its antecedent. These results are interpreted as evidence against a purely syntactic-structure based memory access. However, the exact pattern of effects observed in the data is only partially compatible with the Lewis and Vasishth cue-based parsing model. Therefore, an extension of the Lewis and Vasishth model is proposed. Two principles are added to the original model, namely 'cue confusion' and 'distractor prominence'. Although interference effects are generally interpreted in favor of a content-addressable memory architecture, an alternative explanation for interference effects in reflexive processing has been proposed which, crucially, might reconcile interference effects with a structure-based account. It has been argued that interference effects do not necessarily reflect cue-based retrieval interference in a content-addressable memory but might equally well be accounted for by interference effects which have already occurred at the moment of encoding the antecedent in memory (Dillon, 2011). Three experiments (eye-tracking and self-paced reading) on German reflexives and Swedish possessives were designed to tease apart cue-based retrieval interference from encoding interference. The results of all three experiments suggest that there is no evidence that encoding interference affects the retrieval of a reflexive's antecedent. Taken together, these findings suggest that the processing of reflexives can be explained with the same cue-based retrieval mechanism that has been invoked to explain syntactic dependency resolution in a range of other structures. This supports the view that the language processing system is located within a general cognitive architecture, with a general-purpose content-addressable working memory system operating on linguistic expressions. Finally, two experiments (self-paced reading and eye-tracking) using Chinese relative clauses were conducted to determine the relative contribution to sentence processing difficulty of working-memory processes as compared to syntactic prediction during incremental parsing. Chinese has the cross-linguistically rare property of being a language with subject-verb-object word order and pre-nominal relative clauses. This property leads to opposing predictions of expectation-based accounts and memory-based accounts with respect to the relative processing difficulty of subject vs. object relatives. Previous studies showed contradictory results, which has been attributed to different kinds local ambiguities confounding the materials (Lin and Bever, 2011). The two experiments presented are the first to compare Chinese relatives clauses in syntactically unambiguous contexts. The results of both experiments were consistent with the predictions of the expectation-based account of sentence processing but not with the memory-based account. From these findings, I conclude that any theory of human sentence processing needs to take into account the power of predictive processes unfolding in the human mind.},
  language  = {en}
}
@phdthesis{RoggeSolti2014,
  author    = {Rogge-Solti, Andreas},
  title     = {Probabilistic Estimation of Unobserved Process Events},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus-70426},
  school      = {Universit{\"a}t Potsdam},
  year      = {2014},
  abstract  = {Organizations try to gain competitive advantages, and to increase customer satisfaction. To ensure the quality and efficiency of their business processes, they perform business process management. An important part of process management that happens on the daily operational level is process controlling. A prerequisite of controlling is process monitoring, i.e., keeping track of the performed activities in running process instances. Only by process monitoring can business analysts detect delays and react to deviations from the expected or guaranteed performance of a process instance. To enable monitoring, process events need to be collected from the process environment. When a business process is orchestrated by a process execution engine, monitoring is available for all orchestrated process activities. Many business processes, however, do not lend themselves to automatic orchestration, e.g., because of required freedom of action. This situation is often encountered in hospitals, where most business processes are manually enacted. Hence, in practice it is often inefficient or infeasible to document and monitor every process activity. Additionally, manual process execution and documentation is prone to errors, e.g., documentation of activities can be forgotten. Thus, organizations face the challenge of process events that occur, but are not observed by the monitoring environment. These unobserved process events can serve as basis for operational process decisions, even without exact knowledge of when they happened or when they will happen. An exemplary decision is whether to invest more resources to manage timely completion of a case, anticipating that the process end event will occur too late. This thesis offers means to reason about unobserved process events in a probabilistic way. We address decisive questions of process managers (e.g., "when will the case be finished?", or "when did we perform the activity that we forgot to document?") in this thesis. As main contribution, we introduce an advanced probabilistic model to business process management that is based on a stochastic variant of Petri nets. We present a holistic approach to use the model effectively along the business process lifecycle. Therefore, we provide techniques to discover such models from historical observations, to predict the termination time of processes, and to ensure quality by missing data management. We propose mechanisms to optimize configuration for monitoring and prediction, i.e., to offer guidance in selecting important activities to monitor. An implementation is provided as a proof of concept. For evaluation, we compare the accuracy of the approach with that of state-of-the-art approaches using real process data of a hospital. Additionally, we show its more general applicability in other domains by applying the approach on process data from logistics and finance.},
  language  = {en}
}
@phdthesis{Haider2013,
  author    = {Haider, Peter},
  title     = {Prediction with Mixture Models},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus-69617},
  school      = {Universit{\"a}t Potsdam},
  year      = {2013},
  abstract  = {Learning a model for the relationship between the attributes and the annotated labels of data examples serves two purposes. Firstly, it enables the prediction of the label for examples without annotation. Secondly, the parameters of the model can provide useful insights into the structure of the data. If the data has an inherent partitioned structure, it is natural to mirror this structure in the model. Such mixture models predict by combining the individual predictions generated by the mixture components which correspond to the partitions in the data. Often the partitioned structure is latent, and has to be inferred when learning the mixture model. Directly evaluating the accuracy of the inferred partition structure is, in many cases, impossible because the ground truth cannot be obtained for comparison. However it can be assessed indirectly by measuring the prediction accuracy of the mixture model that arises from it. This thesis addresses the interplay between the improvement of predictive accuracy by uncovering latent cluster structure in data, and further addresses the validation of the estimated structure by measuring the accuracy of the resulting predictive model. In the application of filtering unsolicited emails, the emails in the training set are latently clustered into advertisement campaigns. Uncovering this latent structure allows filtering of future emails with very low false positive rates. In order to model the cluster structure, a Bayesian clustering model for dependent binary features is developed in this thesis. Knowing the clustering of emails into campaigns can also aid in uncovering which emails have been sent on behalf of the same network of captured hosts, so-called botnets. This association of emails to networks is another layer of latent clustering. Uncovering this latent structure allows service providers to further increase the accuracy of email filtering and to effectively defend against distributed denial-of-service attacks. To this end, a discriminative clustering model is derived in this thesis that is based on the graph of observed emails. The partitionings inferred using this model are evaluated through their capacity to predict the campaigns of new emails. Furthermore, when classifying the content of emails, statistical information about the sending server can be valuable. Learning a model that is able to make use of it requires training data that includes server statistics. In order to also use training data where the server statistics are missing, a model that is a mixture over potentially all substitutions thereof is developed. Another application is to predict the navigation behavior of the users of a website. Here, there is no a priori partitioning of the users into clusters, but to understand different usage scenarios and design different layouts for them, imposing a partitioning is necessary. The presented approach simultaneously optimizes the discriminative as well as the predictive power of the clusters. Each model is evaluated on real-world data and compared to baseline methods. The results show that explicitly modeling the assumptions about the latent cluster structure leads to improved predictions compared to the baselines. It is beneficial to incorporate a small number of hyperparameters that can be tuned to yield the best predictions in cases where the prediction accuracy can not be optimized directly.},
  language  = {en}
}
@phdthesis{Zurell2011,
  author    = {Zurell, Damaris},
  title     = {Integrating dynamic and statistical modelling approaches in order to improve predictions for scenarios of environmental change},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus-56845},
  school      = {Universit{\"a}t Potsdam},
  year      = {2011},
  abstract  = {Species respond to environmental change by dynamically adjusting their geographical ranges. Robust predictions of these changes are prerequisites to inform dynamic and sustainable conservation strategies. Correlative species distribution models (SDMs) relate species' occurrence records to prevailing environmental factors to describe the environmental niche. They have been widely applied in global change context as they have comparably low data requirements and allow for rapid assessments of potential future species' distributions. However, due to their static nature, transient responses to environmental change are essentially ignored in SDMs. Furthermore, neither dispersal nor demographic processes and biotic interactions are explicitly incorporated. Therefore, it has often been suggested to link statistical and mechanistic modelling approaches in order to make more realistic predictions of species' distributions for scenarios of environmental change. In this thesis, I present two different ways of such linkage. (i) Mechanistic modelling can act as virtual playground for testing statistical models and allows extensive exploration of specific questions. I promote this 'virtual ecologist' approach as a powerful evaluation framework for testing sampling protocols, analyses and modelling tools. Also, I employ such an approach to systematically assess the effects of transient dynamics and ecological properties and processes on the prediction accuracy of SDMs for climate change projections. That way, relevant mechanisms are identified that shape the species' response to altered environmental conditions and which should hence be considered when trying to project species' distribution through time. (ii) I supplement SDM projections of potential future habitat for black grouse in Switzerland with an individual-based population model. By explicitly considering complex interactions between habitat availability and demographic processes, this allows for a more direct assessment of expected population response to environmental change and associated extinction risks. However, predictions were highly variable across simulations emphasising the need for principal evaluation tools like sensitivity analysis to assess uncertainty and robustness in dynamic range predictions. Furthermore, I identify data coverage of the environmental niche as a likely cause for contrasted range predictions between SDM algorithms. SDMs may fail to make reliable predictions for truncated and edge niches, meaning that portions of the niche are not represented in the data or niche edges coincide with data limits. Overall, my thesis contributes to an improved understanding of uncertainty factors in predictions of range dynamics and presents ways how to deal with these. Finally I provide preliminary guidelines for predictive modelling of dynamic species' response to environmental change, identify key challenges for future research and discuss emerging developments.},
  language  = {en}
}
@phdthesis{GamezLopez2006,
  author    = {G{\´a}mez L{\´o}pez, Antonio Juan},
  title     = {Application of nonlinear dimensionality reduction to climate data for prediction},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus-10956},
  school      = {Universit{\"a}t Potsdam},
  year      = {2006},
  abstract  = {This Thesis was devoted to the study of the coupled system composed by El Ni{\~n}o/Southern Oscillation and the Annual Cycle. More precisely, the work was focused on two main problems: 1. How to separate both oscillations into an affordable model for understanding the behaviour of the whole system. 2. How to model the system in order to achieve a better understanding of the interaction, as well as to predict future states of the system. We focused our efforts in the Sea Surface Temperature equations, considering that atmospheric effects were secondary to the ocean dynamics. The results found may be summarised as follows: 1. Linear methods are not suitable for characterising the dimensionality of the sea surface temperature in the tropical Pacific Ocean. Therefore they do not help to separate the oscillations by themselves. Instead, nonlinear methods of dimensionality reduction are proven to be better in defining a lower limit for the dimensionality of the system as well as in explaining the statistical results in a more physical way [1]. In particular, Isomap, a nonlinear modification of Multidimensional Scaling methods, provides a physically appealing method of decomposing the data, as it substitutes the euclidean distances in the manifold by an approximation of the geodesic distances. We expect that this method could be successfully applied to other oscillatory extended systems and, in particular, to meteorological systems. 2. A three dimensional dynamical system could be modeled, using a backfitting algorithm, for describing the dynamics of the sea surface temperature in the tropical Pacific Ocean. We observed that, although there were few data points available, we could predict future behaviours of the coupled ENSO-Annual Cycle system with an accuracy of less than six months, although the constructed system presented several drawbacks: few data points to input in the backfitting algorithm, untrained model, lack of forcing with external data and simplification using a close system. Anyway, ensemble prediction techniques showed that the prediction skills of the three dimensional time series were as good as those found in much more complex models. This suggests that the climatological system in the tropics is mainly explained by ocean dynamics, while the atmosphere plays a secondary role in the physics of the process. Relevant predictions for short lead times can be made using a low dimensional system, despite its simplicity. The analysis of the SST data suggests that nonlinear interaction between the oscillations is small, and that noise plays a secondary role in the fundamental dynamics of the oscillations [2]. A global view of the work shows a general procedure to face modeling of climatological systems. First, we should find a suitable method of either linear or nonlinear dimensionality reduction. Then, low dimensional time series could be extracted out of the method applied. Finally, a low dimensional model could be found using a backfitting algorithm in order to predict future states of the system.},
  subject      = {Nichtlineare Dynamik},
  language  = {en}
}