SSIAB 12

Wednesday, May 23

08:30 - 09:00 Welcome

09:00 - 09:25 Andreas Dyreborg Christoffersen (Aalborg Univ. Denmark)
Pair correlation functions and limiting distributions of iterated cluster point processes

We consider a Markov chain of point processes such that each state is a super position of an independent cluster process with the previous state as its centre process together with some independent noise process. The model extends earlier work by Felsenstein and Shimatani describing a reproducing population. We discuss when closed term expressions of the first and second order moments are available for a given state. In a special case it is known that the pair correlation function for these type of point processes converges as the Markov chain progresses, but it has not been shown whether the Markov chain has an equilibrium distribution with this, particular, pair correlation function and how it may be constructed. Assuming the same reproducing system, we construct an equilibrium distribution by a coupling argument.
J. Møller and A.D. Christoffersen (2017). Pair correlation functions and limiting distributions of iterated cluster point processes. Submitted for journal publication. Available at arXiv: 1711.08984. Research Report 11, 2017, Centre for Stochastic Geometry and Advanced Bioimaging.

09:25 - 09:50 Clæs Andersson (Chalmers University, Gothenburg)
A cluster process with a flexible cluster size distribution, and application to nerve fiber patterns

In a previous study we used the Thomas process to model the entry points of epidermal nerve fibers (ENFs), and the results motivated us to consider a model with a more flexible cluster size distribution. To achieve this we modify the Thomas process and let the cluster sizes follow a generalized Poisson distribution, which has an extra parameter allowing for over- and under-dispersion of the distribution.
Although simple in its construction, this model presents certain challenges when it comes to parameter estimation. One is that the intensity and the second order properties are not enough to identify the model, meaning that the minimum contrast method based on the K- or pair correlation-function is not directly applicable. Another challenge is that this process is in general not a Cox process, meaning that MCMC estimation is more complicated than for the Thomas process.
In this work, a minimum contrast method and a Bayesian MCMC method for estimating the parameters of the model are presented and evaluated in a simulation study. Moreover, the model is fitted to ENF entry point patterns, and the results clearly indicate that the cluster sizes are under-dispersed.
This is joint work with Tomáš Mrkvička.

09:50 - 10:15 Viktor Benes (Charles University, Prague)
Modelling and simulation of polycrystalline materials using 3D tessellations

In this talk we investigate three-dimensional tessellations and their application to the modelling of microstructures of polycrystalline materials. Variety of tessellation models are fit to voxelized microstructural specimens by optimizing a discrepancy measure using the simulated annealing. Characteristics of the fit are evaluated. Then the problem of model selection is tried to be solved using some methods from statistical learning theory, among them the structural risk minimization. Numerical results are presented on both simulated and real data.
This is a joint work with O. Šedivý, J. Kopeček from Prague and D. Westhoff, C.E. Krill III, V. Schmidt from Ulm.

---

10:15 - 10:45 Break

---

10:45 - 11:10 Heidi Søgaard Christensen (Aalborg University)
Structured space-sphere point processes and K-functions

11:10 - 11:35 Mari Myllymäki (Natural Resources Institute Finland)
A one-way ANOVA test for functional data with graphical interpretation

We introduce a new one-way functional ANOVA test based on global envelope tests (Myllymäki et al., 2017, Global envelope tests for spatial processes. J. R. Stat. Soc. B 79, 381-404, 2017). The advantage of the proposed test is that it identifies the domains of the functions which are responsible for the potential rejection of the equality of means of the groups. We introduce two versions of this test: the first gives a graphical interpretation of the test results in the original space of the functions and the second immediately offers a post-hoc test by identifying the significant pair-wise differences between groups. The new test is supported by the R library GET (https://github.com/myllym/GET).

11:35 - 12:00 Tomáš Mrkvička (University of South Bohemia)
Refinements of the global envelope tests, with application on the General linear model of neuroimage data

The global envelope tests defined in Myllymäki et al. (2017) are computational tools for graphical testing of a complex null hypothesis where the testing is based on the functional summary statistic. When the summary function is discredited in many points, the methods can achieve many ties in ordering of summary functions obtained from simulation of the null model. Here we study three different refinement of the global envelope test which solve this problem.
The refinements are applied on the general linear model for d-dimensional functional data or image data. Especially, on the permutation methods, which require only minimum assumptions. Such analyze consists of several crucial steps. First the appropriate test statistic computed for each single voxel has to be chosen such that it is informative about the studied null and alternative hypothesis and which is homogeneous across the space. Second the appropriate permutation scheme has to be used to generate the permutations from studied null hypothesis. Third appropriate multiple testing correction has to be applied on significance results obtained for all voxels by permutation test. Here the third problem is considered. Especially, the new methods for multiple testing are introduced in this problem by using refinements of global envelope tests. Their performances are analyzed through simulation study of general linear models of image data and they are compared to the other multiple testing approaches.
Myllymäki M., Mrkvička T., Seijo H., Grabarnik P., Hahn U.: Global envelope tests for spatial processes, JRSS Series B 79/2, 2017, 381-404.

---

12:00 - 14:00 Lunch

---

14:00 - 14.25 Juha Heikkinen (Luke, Finland)
Detecting long-term effects in animal movement

This is joint work with Anna-Kaisa Ylitalo (1,2) and Ilpo Kojola (1) (1-Natural Resources Institute Finland and 2-University of Jyväskylä).
One of the topical issues in animal movement research is the influence of spatial cognition and memory. There are still very few studies addressing it with a statistical analysis of empirical data (Schlägel et al. 2017). This presentation outlines ongoing work aimed at unraveling long-term patterns in the movements of grey wolf (Canis lupus) individuals within their established home range.
The data set contains GPS telemetry locations of about 50 wolves with approximately one year's monitoring of each individual and mainly 4 hours' time interval between the fixes. For model development, we also have about two months' intensive surveillance data on 10 wolves with 30 minutes' time interval and field checks of visited locations.
This talk will focus on the analysis of intensive surveillance data from summer months, when the pups cannot yet move very far away from the den. The movement of both alpha female and male is then characterized by daily returns to the den. The main motivations for trips away from the den are thought to be hunting, returns to carcasses or hides, and territorial patrolling.
We present some preliminary findings concerning the space-time patterns in the trips, and some challenges in their spatio-temporal modelling.
Schlägel, U. E. Merrill, E. H., and Lewis, M. A. 2017. Territory surveillance and prey management: Wolves keep track of space and time. Ecology and Evolution 7:838-8405.

14:25 - 14:50 Svenia Behm (University of Passau)
Additive semiparametric land use regression

Existing land use regression (LUR) approaches usually employ parametric assumptions to model the conditional distribution of air pollutant measurements. We propose a flexible data-driven additive semi-parametric framework for modeling the annual mean nitrogen dioxide concentration across Germany as measured by background monitoring sites. Predictor variables are longitude, latitude, altitude, population density, land use and road traffic intensity. Resting on the crucial assumption of additivity of the structural and spatial model components, the structural predictors enter the model either linearly or via univariate splines and the spatial component enters the model either linearly or via a bivariate spline.
Using leave-one-out-cross-validation (LOOCV) in- and out-of-sample statistics, we find that a pre-selection of the predictor variables reduces the mean absolute error (MAE), the root mean square error (RMSE) and the information criteria BIC and AIC. Through exploratory analysis, we find considerable spatial variation and nonlinearities in the data. We therefore model population density via a univariate spline, the spatial component via a bivariate spline and the other predictors in linear additive form. In- and out-of-sample metrics support the proposed model.
Additive semi-parametric models are a promising choice to analyse and predict the conditional distribution of air pollutant concentration. Our specification allows to account for local heterogeneity, potential nonlinearities and spatial anisotropy in a flexible, data-driven way -- while avoiding to impose parametric assumptions a priori (i.e., without looking at the data). A straightforward extension of our approach is to model several characteristics of the conditional pollutant distribution such as quantiles or expectiles.
This is joint work with Harry Haupt and Markus Fritsch.

14:50 - 15:15 Aila Särkkä (Chalmers University, Gothenburg)
Characterizing cross-subject spatial interaction patterns in functional magnetic resonance imaging studies

Conventionally, multi-subject functional Magnetic Resonance Imaging (fMRI) methods rely on combining information across subjects one voxel at a time in order to identify locations of peak activation in the brain. We have developed a two-stage model which addresses shortcomings of standard methods by explicitly modelling the spatial structure of functional signals and recognizing that corresponding cross-subject functional signals can be spatially misaligned. In our first stage analysis, we introduce a marked spatial point process model that captures the spatial features of the functional response and identifies a configuration of activation units for each subject. The locations of these activation units are used as input for the second stage model. The point process model of the second stage analysis is developed to characterize multi-subject activation patterns by estimating the strength of cross-subject interactions at different spatial ranges. The model is a modification of Geyer’s saturation model. We applied our methods to an fMRI study of 21 individuals who performed an attention test. We identified four brain regions that are involved in the test and found that our model results agree well with our understanding of how these regions engage with the tasks performed during the attention test. Our results highlighted that cross-subject interactions are stronger in brain areas that have a more specific function in performing the experimental tasks than in other areas.
Joint work with Adél Lee (Etosha Business and Research Consulting, GA, USA), Tara M. Madhyastha (University of Washington, USA), and Thomas J. Grabowski (University of Washington, USA).

---

15:15 - 16:00 Coffee Break

---

16:00 - 16:25 Jacob Gulddahl Rasmussen (Aalborg Univ. Denmark)
Isotropic covariance functions on graphs and their edges

We develop parametric classes of covariance functions on linear networks and their extension to graphs with Euclidean edges, i.e., graphs with edges viewed as line or curve segments allowing us to consider points on the graph which are either vertices or points on an edge. The covariance functions are defined on the vertices and edge points of these graphs and are isotropic in the sense that they depend only on the geodesic distance or on a new metric called the 3resistance metric (corresponding to resistance in an electrical network). We discuss the advantages of using the resistance metric in comparison with the geodesic metric as well as the restrictions these metrics impose on the investigated covariance functions. In particular, many of the commonly used isotropic covariance functions in the spatial statistics literature (the power exponential, Matérn, generalized Cauchy, and Dagum classes) are shown to be valid with respect to the resistance metric for any graph with Euclidean edges, whilst they are only valid with respect to the geodesic metric in more special cases. This is joint work with Ethan Anderes and Jesper Møller.

16:25 - 16:50 Nicolas Chenavier (Université du Littoral Cote d'Opale,Calais)
The maximal degree in a Poisson-Delaunay graph

16:50 - 17:15 Marie-Colette van Lieshout (CWI, Amsterdam)
Nearest-neighbour Markov point processes on graphs with Euclidean edges

We define nearest-neighbour point processes on graphs with Euclidean edges and linear networks. They can be seen as the analogues of renewal processes on the real line. We show that the Delaunay neighbourhood relation on a tree satisfies the Baddeley-Møller consistency conditions and provide a characterisation of Markov functions with respect to this relation. We show that a modified relation defined in terms of the local geometry of the graph satisfies the consistency conditions for all graphs with Euclidean edges.

---

Thursday, May 24

09:00 - 09:25 Jean-François Coeurjolly (UQAM, Montreal)
The median of a jittered Poisson distribution is close to ...

09:25 - 09:50 Daniela Flimmel Novotná (Charles University, Prague)
Unbiased estimators of weighted Voronoi cell characteristics

Consider a unit stationary independently marked Poisson point process on the Euclidean space of a general dimension with marks in the space of nonnegative real numbers. We deal with three types of tessellation generated by this process: the Voronoi tessellation, the Laguerre tessellation and the additively weighted model. We observe only those cells that are fully contained in a given observation window and study limit behaviour of geometric statistics of these cells while the window increases to the whole Euclidean space. The geometric statistics of interest are of the Horvitz-Thompson type (a sum of scores evaluated in each cell). When divided by the volume of the window, we obtain an unbiased estimator of the expected value of the score in the typical cell. It can be shown that the scores are exponentially stabilizing under each considered type of tessellation and hence with some additional moment assumptions, one can prove the asymptotic normality using stabilization methods.
This is a joint work with Joseph E. Yukich and Zbynek Pawlas.

09:50 - 10:15 Thomas Opitz (INRA, Avignon)
Flexible spatial process models based on Lévy indicator convolutions

Process convolutions yield flexible stochastic processes beyond the realm of Gaussianity, but statistical inference is often hampered by the lack of closed-form marginal distributions. We here remedy such issues through a class of process convolutions based on smoothing a (d+1)-dimensional Lévy basis with an indicator function kernel to construct a d-dimensional convolution process. Indicator kernels ensure univariate distributions in the Lévy basis family (such as gamma, Gumbel, inverse gaussian, stable, student, Poisson, negative binomial, and many others), which provides a sound basis for interpretation, parametric modeling and statistical estimation. We propose a class of stationary and isotropic convolution processes constructed through hypograph indicator sets defined as the space between the curve (s,H(s)) of a spherical probability density function H and the plane (s,0). If H is radially nonincreasing, the covariance is expressed through the univariate distribution function of H. The bivariate joint tail behavior in such convolution processes and some interesting links to extreme value theory will be explored. For statistical inference of parametric models, we develop pairwise likelihood techniques. This modeling framework is illustrated on a real data example.

---

10:15 - 10:45 Coffee Break

---

10:45 - 11:10 Henrike Häbel (Natural Resources Institute, Finland)
Detection and classification of the spatial structure of trees with remote sensing

Forest biodiversity can be divided into three major aspects, namely spatial distribution, species diversity, and variation in attributes. Biodiversity has been assessed by considering indices independent or dependent on distances between neighboring trees at stand-level. Examples are Pielou’s segregation index and the species mingling index. With recent developments in remote sensing technologies, it becomes intriguing to study to what extent biodiversity indices can be estimated from remote sensing data. In my talk, I will focus on the detection and classification of the spatial structure of trees based on airborne laser scanning (ALS) data from Finnish forests. I will show how summary characteristics from spatial statistics can be useful for a spatial analysis of biodiversity indices.
This is joined work with Mari Myllymäki and András Balázs.

11:10 - 11:35 Lauri Mehtätalo (University of Eastern, Finland)
Finding the hidden trees in remote sensing: could point patterns and stochastic geometry provide a solution?

Tree canopies can be seen as 3-dimensional random sets, and the forest canopy of a certain area is the union of these. Airborne laser scanners provide measurements of the forest canopy, and the parameters of interest for forest managers are the individual trees that form this union. Stochastic geometry analyzes random sets that are unions of random closed sets, and is therefore a natural starting point for such analysis. This talk summarizes the past works on the use of stochastic geometry in estimating forest characteristic of interest using area-level canopy data. The special focus is in the forest inventories based on aerial laser scanning data.

11:35 - 12:00 Mikko Kuronen (Natural Resources Institute, Finland)
Bayesian inference for spatial tree regeneration models using MCMC and INLA

---

12:00 - 14:00 Lunch

---

14:00 - 14.25 Arnaud Poinas (University of Rennes 1)
Mixing properties and CLT for determinantal point processes

In this talk we focus on the negative association property of determinantal point process (DPP for short) and its consequences. Negative association is a property that characterizes the negative dependency of a stochastic process. Well studied in random field theory, this notion rarely appears in the point process literature. We show that negative association implies alpha-mixing properties as well as a general CLT, stronger than classical CLT based on alpha-mixing. DPPs being negatively associated, we derive a CLT for a wide class of functionnals of non-stationary DPPs, that include the type of statistics involved in asymptotic inference of these processes.

14:25 - 14:50 Lavancier Frederic (University of Nantes)
Adaptive estimating function inference for inhomogeneous determinantal point processes

14:50 - 15:15 Adrien Mazoyer (UQAM, Montreal)
The use of Determinantal Point Processes for computer experiments

---

15:15 - 16:00 Coffee Break

---

16:00 - 16:25 Achmad Choiruddin (Aalborg Univ. Denmark)
Sparse models for highly multivariate log-Gaussian Cox processes

Based on the joint-work with Rasmus Waagepetersen, Jean-François Coeurjolly and Francisco Cuevas-Pachecho
Studies regarding biodiversity in tropical rainforest ecology are conducted using large data sets containing locations of thousands of trees for each of hundreds of species. To get insight in the multivariate dependence structure for a high number of species, a model-based approach is required. Waagepetersen et. al., (2016) proposed to use multivariate log-Gaussian Cox process models but their data analysis only involved nine species. One problem with using their approach for a large number of species is model complexity where the numbers of parameters increase very fast as a function of the number of species. Another related problem is computational where general off-the-shelf optimization algorithms may be slow and unstable partly due to the high dimension of the parameter space.
In this project we aim at extending the methodology in Waagepetersen et. al., (2016) by introducing regularization of parameter estimates and constructing more reliable optimization methods. In addition to providing computationally more stable results, use of Lasso regularization may also lead to biologically more interpretable models. To do so, we note that the estimation method from Waagepetersen et. al., (2016) can be approximated by a least squares method which enables us to benefit from well-known and robust Lasso and elastic net techniques.
Reference: Waagepetersen, R., Guan, Y., Jalilian, A., & Mateu, J. (2016). Analysis of multispecies point patterns by using multivariate log‐Gaussian Cox processes. Journal of the Royal Statistical Society: Series C (Applied Statistics), 65(1), 77-96.

16:25 - 16:50 Ottmar Cronie (Umeå University)
Resample-smoothing of intensity estimators

Voronoi intensity estimators, which are non-parametric estimators for intensity functions of point processes, are both parameter-free and adaptive. Their major drawback, however, is that they tend to under-smooth the data, in particular in places where the point density of the observed point pattern is high. With the aim of improving the performance of Voronoi intensity estimators, we propose an additional smoothing technique for intensity estimators for point processes, which is based on repeated independent thinnings of the point process/pattern. We show unbiasedness and variance results, propose a rule-of-thumb and a data-driven approach to choosing the amount of smoothing to apply, and evaluate our approach numerically for planar point processes. We finally apply our proposed intensity estimation scheme to a linear network dataset.
This is joint work with M. Moradi, E. Rubak, A. Baddeley and J. Mateu.

16:50 - 17:15 Abdollah Jalilian (Razi University, Kermanshah)
Estimation of the pair correlation function

---

18:30 - 19:00 Administrative meeting Future of the workshop, next organisers ...

19:00 Conference Dinner Menu

Starter: Cod Fish, Norway lobster, hand-peeled shrimp with garlic sauce and salads
Main course: Grilled organic pork filet, long fried pork breast with butter-fried white asparagus, and potatoes with parsley (NB:Please, let us known as soon as possible if you are interested in a vegetarian alternative)
Dessert: Chocolate cake with vanilla ice cream

---

Friday, May 25

09:00 - 09:25 Zbynek Pawlas (Charles University, Prague)
Statistical inference for random marked closed sets

By a random marked closed set we understand a random upper semi-continuous function on a random domain (random closed set). Special examples are marked point processes and random fields (the corresponding random closed set is a point process and some deterministic domain, respectively). It is possible to test the null hypothesis of independence between the random domain and the random mark function. In practical applications, it is common that these components are dependent (mark-set interactions are present). Summary characteristics for stationary random marked closed sets describe specific aspects of the distribution of both components and the mark-set interactions. The most popular are second-order characteristics (e.g. mark weighted K-function and mark correlation function). We are interested in their non-parametric estimation. Some of the statistical properties of the estimators will be presented.

09:25 - 09:50 Katerina Helisova (Czech Technical University, Prague)
Similarity of random sets based on approximations by convex compact sets and envelope tests

09:50 - 10:15 Vesna Gotovac (University of Split)
Similarity measures of random sets based on N-distances and their applications to two-realisation problem

General random sets could be of very ragged shapes and therefore it is difficult to describe them by simple models and compare them. In this contribution, we focus on statistical testing of similarity of random sets in a non-parametric way. The talk presents some types of similarity measures of random sets based on just two realisations of each set. Such an approach has been developed for germ-grain type of random sets models. When the grains are hard to distinguish, we compare their inner structure with the aim to capture the repulsion or clustering tendencies of the set components. When the grains are easily distinguishable and ragged shaped, the approach takes into account the general position, number and the shape of the grains in order to obtain similarity between two realisations. All the similarity measures are given by the p-values of tests of equality in distribution based on N-distances. Both approaches are justified by a simulation study and applied to real data.

---

10:15 - 10:45 Coffee Break

---

10:45 - 11:10 Jiri Dvorak (Charles University, Prague)
Stochastic reconstruction for inhomogeneous point patterns

In statistical analysis of point patterns we often use simulation-based Monte Carlo tests in cases where the null distribution of the test statistic cannot be described analytically. This often happens if we use functional test statistics such as the K-function. However, we often need to test hypotheses which are not specific enough to allow simulation required to perform a classical Monte Carlo test. This is the case of hypotheses such as ``isotropy of the point process'' or ``independence in the multi-type point process''.
In such situation we can take advantage of the stochastic reconstruction algorithm [1,2] which produces point patterns with similar properties as the observed pattern, without the need to specify a particular point process model and estimate its parameters.
The algorithm of stochastic reconstruction for stationary point processes was suggested in [2] and the case of point processes with non-constant intensity function was discussed in [3].
We give a brief overview of the stochastic reconstruction algorithm and discuss that the approach of [3] needs to be modified in order to produce outputs with the correct intensity function. We present such a modified algorithm and demonstrate, using simulated datasets, that both the intensity function and the interactions between points are correctly reconstructed. This is a joint work with Kateřina Koňasová.
[1] S. Torquato. Random Heterogeneous Materials. Microstructure and Macroscopic Properties. Springer-Verlag, New York, 2002.
[2] A. Tscheschel and D. Stoyan. Statistical reconstruction of random point patterns. Computational Statistics & Data Analysis 51(2), 859--871, 2006.
[3] T. Wiegand, F. He and S.P. Hubbell. A systematic comparison of summary characteristics for quantifying point patterns in ecology. Ecography 36(1), 92--103, 2013.

11:10 - 11:35 Mohammad Mehdi Moradi (University Jaume I of Castellon)
Trajectories: Classes and Methods for Trajectory Data

11:35 - 12:00 Tuomas Rajala (University College London)
Spatial inter-species interactions in plant communities: Independence, or lack of data?

---

12:00 - 14:00 Lunch

---

14:00 - End of the conference - Good travel or Enjoy Aalborg Carnival

List of speakers

• Heidi Søgaard Christensen (Aalborg University)
• Andreas Dyreborg Christoffersen (Aalborg University)
• Jakob Rasmussen (Aalborg University)
• Achmad Choiruddin (Aalborg University)
• Lauri Mehtätalo (University of Eastern Finland)
• Mari Myllymäki (Natural Resources Institute Finland)
• Mikko Kuronen (Natural Resources Institute, Finland)
• Henrike Häbel (Natural Resources Institute, Finland)
• Juha Heikkinen (Luke, Finland)
• Viktor Benes (Charles University, Prague)
• Daniela Flimmel Novotná (Charles University, Prague)
• Zbynek Pawlas (Charles University, Prague)
• Jiri Dvorak (Charles University, Prague)
• Aila Särkkä (Chalmers University, Gothenburg)
• Clæs Andersson (Chalmers University, Gothenburg)
• Mohammad Mehdi Moradi (University Jaume I of Castellon)
• Nicolas Chenavier (Université du Littoral Cote d'Opale, Calais)
• Thomas Opitz (INRA, Avignon)
• Jalilian Abdollah (Razi University, Kermanshah)
• Jean-François Coeurjolly (UQAM, Montreal)
• Adrien Mazoyer (UQAM, Montreal)
• Tomas Mrkvicka (University of South Bohemia)
• Marie-Colette van Lieshout (CWI, Amsterdam)
• Vesna Gotovac (University of Split)
• Katerina Helisova (Czech Technical University, Prague)
• Ottmar Cronie (Umeå University)
• Svenia Behm (University of Passau)
• Tuomas Rajala (University college London)
• Arnaud Poinas (University of Rennes 1)
• Frédéric Lavancier (University of Nantes)

The workshop is held at Comwell Hvide Hus Hotel, Vesterbro 2, 9000 Aalborg, Denmark.

The city is well-served by plane and train.

From the airport

By bus (approx. 30min): The bus stop is located in front of the airport. Take bus number 12 in direction of Aalborg Universitet and stop at Aalborg Busterminal. You may see a letter near the number 12 but it is not relevant. Then take the underpass and cross the small park. The place of the conference is the only tower visible.

By taxi (approx. 15min): Ask for Comwell hotel.

From the train station

Stop at Aalborg St. Go out of the station and go through the underpass to reach a park. The Comwell Hotel is the tower on the other side of the park.

By car:

Head north on the E45 and take the exist 28-Aalborg S toward Route 180 in direction of the hospital. The Comwell Hotel lay just after the first intersection after the hospital.

Please fill the following form.

Conference Dinner

A conference dinner will be held on the 24th May.

Aalborg Carneval

The day after the workshop (i.e. Saturday, May 26) there is carneval in Aalborg (actually North Europe's biggest carneval), so you may wish to stay for weekend if you like to see this. Also there may be some carneval activities around the city center in the evenings on the days of the workshop. For details, see this link.

Important note: financial support to attend the conference may cover hotel expense up to Friday 25 May.