``` r
ggraph(pa,layout="stress")+
geom_edge_link(width=0.2,colour="grey")+
geom_node_point(col="black",size=0.3)+
theme_graph()
```
## Stress Majorization: Unconnected Network
Stress majorization also works for networks with several components. It
relies on a bin packing algorithm to efficiently put the components in a
rectangle, rather than a circle.
``` r
set.seed(666)
g <- disjoint_union(
sample_pa(10,directed = F),
sample_pa(20,directed = F),
sample_pa(30,directed = F),
sample_pa(40,directed = F),
sample_pa(50,directed = F),
sample_pa(60,directed = F),
sample_pa(80,directed = F)
)
ggraph(g,layout = "nicely") +
geom_edge_link() +
geom_node_point() +
theme_graph()
```
``` r
ggraph(g, layout = "stress",bbox = 40) +
geom_edge_link() +
geom_node_point() +
theme_graph()
```
## Backbone Layout
Backbone layouts are helpful for drawing hairballs.
``` r
set.seed(665)
#create network with a group structure
g <- sample_islands(9,40,0.4,15)
g <- simplify(g)
V(g)$grp <- as.character(rep(1:9,each=40))
ggraph(g,layout = "stress")+
geom_edge_link(colour=rgb(0,0,0,0.5),width=0.1)+
geom_node_point(aes(col=grp))+
scale_color_brewer(palette = "Set1")+
theme_graph()+
theme(legend.position = "none")
```
The backbone layout helps to uncover potential group structures based on
edge embeddedness and puts more emphasis on this structure in the
layout.
``` r
bb <- layout_as_backbone(g,keep=0.4)
E(g)$col <- F
E(g)$col[bb$backbone] <- T
ggraph(g,layout="manual",x=bb$xy[,1],y=bb$xy[,2])+
geom_edge_link(aes(col=col),width=0.1)+
geom_node_point(aes(col=grp))+
scale_color_brewer(palette = "Set1")+
scale_edge_color_manual(values=c(rgb(0,0,0,0.3),rgb(0,0,0,1)))+
theme_graph()+
theme(legend.position = "none")
```
## Radial Layout with Focal Node
The function `layout_with_focus()` creates a radial layout around a
focal node. All nodes with the same distance from the focal node are on
the same circle.
``` r
library(igraphdata)
library(patchwork)
data("karate")
p1 <- ggraph(karate,layout = "focus",focus = 1) +
draw_circle(use = "focus",max.circle = 3)+
geom_edge_link(edge_color="black",edge_width=0.3)+
geom_node_point(aes(fill=as.factor(Faction)),size=2,shape=21)+
scale_fill_manual(values=c("#8B2323", "#EEAD0E"))+
theme_graph()+
theme(legend.position = "none")+
coord_fixed()+
labs(title= "Focus on Mr. Hi")
p2 <- ggraph(karate,layout = "focus",focus = 34) +
draw_circle(use = "focus",max.circle = 4)+
geom_edge_link(edge_color="black",edge_width=0.3)+
geom_node_point(aes(fill=as.factor(Faction)),size=2,shape=21)+
scale_fill_manual(values=c("#8B2323", "#EEAD0E"))+
theme_graph()+
theme(legend.position = "none")+
coord_fixed()+
labs(title= "Focus on John A.")
p1+p2
```
## Radial Centrality Layout
The function `layout_with_centrality` creates a radial layout around the
node with the highest centrality value. The further outside a node is,
the more peripheral it is.
``` r
library(igraphdata)
library(patchwork)
data("karate")
bc <- betweenness(karate)
p1 <- ggraph(karate,layout = "centrality", centrality = bc, tseq = seq(0,1,0.15)) +
draw_circle(use = "cent") +
annotate_circle(bc,format="",pos="bottom") +
geom_edge_link(edge_color="black",edge_width=0.3)+
geom_node_point(aes(fill=as.factor(Faction)),size=2,shape=21)+
scale_fill_manual(values=c("#8B2323", "#EEAD0E"))+
theme_graph()+
theme(legend.position = "none")+
coord_fixed()+
labs(title="betweenness centrality")
cc <- closeness(karate)
p2 <- ggraph(karate,layout = "centrality", centrality = cc, tseq = seq(0,1,0.2)) +
draw_circle(use = "cent") +
annotate_circle(cc,format="scientific",pos="bottom") +
geom_edge_link(edge_color="black",edge_width=0.3)+
geom_node_point(aes(fill=as.factor(Faction)),size=2,shape=21)+
scale_fill_manual(values=c("#8B2323", "#EEAD0E"))+
theme_graph()+
theme(legend.position = "none")+
coord_fixed()+
labs(title="closeness centrality")
p1+p2
```
## Large graphs
`graphlayouts` implements two algorithms for visualizing large networks
(\<100k nodes). `layout_with_pmds()` is similar to `layout_with_mds()`
but performs the multidimensional scaling only with a small number of
pivot nodes. Usually, 50-100 are enough to obtain similar results to the
full MDS.
`layout_with_sparse_stress()` performs stress majorization only with a
small number of pivots (~50-100). The runtime performance is inferior to
pivotMDS but the quality is far superior.
A comparison of runtimes and layout quality can be found in the
[wiki](https://github.com/schochastics/graphlayouts/wiki/)
**tl;dr**: both layout algorithms appear to be faster than the fastest
igraph algorithm `layout_with_drl()`.
Below are two examples of layouts generated for large graphs using
`layout_with_sparse_stress()`
A retweet network with ~18k nodes and ~61k edges (runtime:
45.2s)
A co-citation network with ~12k nodes and ~68k edges (runtime: 21s)
## dynamic layouts
`layout_as_dynamic()` allows you to visualize snapshots of longitudinal
network data. Nodes are anchored with a reference layout and only moved
slightly in each wave depending on deleted/added edges. In this way, it
is easy to track down specific nodes throughout time. Use `patchwork` to
put the individual plots next to each other.
``` r
library(patchwork)
#gList is a list of longitudinal networks.
xy <- layout_as_dynamic(gList,alpha = 0.2)
pList <- vector("list",length(gList))
for(i in 1:length(gList)){
pList[[i]] <- ggraph(gList[[i]],layout="manual",x=xy[[i]][,1],y=xy[[i]][,2])+
geom_edge_link0(edge_width=0.6,edge_colour="grey66")+
geom_node_point(shape=21,aes(fill=smoking),size=3)+
geom_node_text(aes(label=1:50),repel = T)+
scale_fill_manual(values=c("forestgreen","grey25","firebrick"),guide=ifelse(i!=2,FALSE,"legend"))+
theme_graph()+
theme(legend.position="bottom")+
labs(title=paste0("Wave ",i))
}
Reduce("+",pList)+
plot_annotation(title="Friendship network",theme = theme(title = element_text(family="Arial Narrow",face = "bold",size=16)))
```
## Layout manipulation
The functions `layout_mirror()` and `layout_rotate()` can be used to
manipulate an existing
layout
graphlayouts/man/ 0000755 0001762 0000144 00000000000 13526724506 013561 5 ustar ligges users graphlayouts/man/draw_circle.Rd 0000644 0001762 0000144 00000001565 13526724506 016335 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/annotate_functions.R
\name{draw_circle}
\alias{draw_circle}
\title{Draw concentric circles}
\usage{
draw_circle(col = "#00BFFF", use = "focus", max.circle)
}
\arguments{
\item{col}{color of circles}
\item{use}{one of 'focus' or 'cent'}
\item{max.circle}{if use = 'focus' specifies the number of circles to draw}
}
\value{
concentric circles around origin
}
\description{
Draw concentric circles
}
\details{
this function is best used with a concentric layout such as \link{layout_with_focus} and \link{layout_with_centrality}.
}
\examples{
library(igraph)
library(ggraph)
g <- sample_gnp(10,0.4)
\dontrun{
ggraph(g,layout = "centrality",centrality = degree(g))+
draw_circle(use = "cent")+
geom_edge_link()+
geom_node_point(shape = 21,fill = "grey25",size = 5)+
theme_graph()+
coord_fixed()
}
}
graphlayouts/man/graph_manipulate.Rd 0000644 0001762 0000144 00000001364 13526724506 017374 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/graph_manipulate.R
\name{graph_manipulate}
\alias{graph_manipulate}
\alias{reorder_edges}
\title{Manipulate graph}
\usage{
reorder_edges(g, attr, desc = TRUE)
}
\arguments{
\item{g}{igraph object}
\item{attr}{edge attribute name used to sort edges}
\item{desc}{logical. sort in descending (default) or ascending order}
}
\value{
manipulated graph
}
\description{
functions to manipulate a graph
}
\details{
\code{reorder_edges()} allows to reorder edges according to an attribute so that edges are
drawn in the given order.
}
\examples{
library(igraph)
library(ggraph)
g <- sample_gnp(10,0.5)
E(g)$attr <- 1:ecount(g)
gn <- reorder_edges(g,"attr")
}
\author{
David Schoch
}
graphlayouts/man/layout_dynamic.Rd 0000644 0001762 0000144 00000003537 13526724655 017106 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/layout_dynamic.R
\name{layout_dynamic}
\alias{layout_dynamic}
\alias{layout_as_dynamic}
\title{dynamic graph layout}
\usage{
layout_as_dynamic(gList, weights = NA, alpha = 0.5, iter = 500,
tol = 1e-04)
}
\arguments{
\item{gList}{list of igraph objects. Each network must contain the same set of nodes.}
\item{weights}{possibly a numeric vector with edge weights. If this is NULL and the graph has a weight edge attribute, then the attribute is used. If this is NA then no weights are used (even if the graph has a weight attribute). By default, weights are ignored. See details for more.}
\item{alpha}{weighting of reference layout. See details.}
\item{iter}{number of iterations during stress optimization}
\item{tol}{stopping criterion for stress optimization}
}
\value{
list of coordinates for each graph
}
\description{
Create layouts for longitudinal networks.
}
\details{
The reference layout is calculated based on the union of all graphs. The parameter alpha controls the influence of the reference layout.
For alpha=1, only the reference layout is used and all graphs have the same layout. For alpha=0, the stress layout of each individual graph is used. Values in-between interpolate between the two layouts.
Be careful when using weights. In most cases, the inverse of the edge weights should be used to ensure that the endpoints of an edges with higher weights are closer together (weights=1/E(g)$weight).
}
\examples{
library(igraph)
g1 <- sample_gnp(20,0.2)
g2 <- sample_gnp(20,0.2)
g3 <- sample_gnp(20,0.2)
xy <- layout_as_dynamic(list(g1,g2,g3))
# layout for first network
xy[[1]]
}
\references{
Brandes, U. and Indlekofer, N. and Mader, M. (2012). Visualization methods for longitudinal social networks and stochastic actor-oriented modeling. \emph{Social Networks} 34 (3) 291-308
}
graphlayouts/man/layout_backbone.Rd 0000644 0001762 0000144 00000002765 13526724506 017223 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/layout_backbone.R, R/layouts.R
\name{layout_backbone}
\alias{layout_backbone}
\alias{layout_as_backbone}
\alias{layout_igraph_backbone}
\title{backbone graph layout}
\usage{
layout_as_backbone(g, keep = 0.2, backbone = TRUE)
layout_igraph_backbone(g, keep = 0.2, backbone = TRUE, circular)
}
\arguments{
\item{g}{igraph object}
\item{keep}{fraction of edges to keep during backbone calculation}
\item{backbone}{logical. Return edge ids of the backbone (Default: TRUE)}
\item{circular}{not used}
}
\value{
list of xy coordinates and vector of edge ids included in the backbone
}
\description{
emphasizes a hidden group structure if it exists in the graph. Calculates a layout for a sparsified network only including the most embedded edges. Deleted edges are added back after the layout is calculated.
}
\details{
The layout_igraph_* function should not be used directly. It is only used as an argument for plotting with 'igraph'.
'ggraph' natively supports the layout.
}
\examples{
library(igraph)
g <- sample_islands(9,20,0.4,9)
g <- simplify(g)
V(g)$grp <- as.character(rep(1:9,each=20))
bb <- layout_as_backbone(g,keep=0.4)
# add backbone links as edge attribute
E(g)$col <- FALSE
E(g)$col[bb$backbone] <- TRUE
}
\references{
Nocaj, A., Ortmann, M., & Brandes, U. (2015). Untangling the hairballs of multi-centered, small-world online social media networks. Journal of Graph Algorithms and Applications: JGAA, 19(2), 595-618.
}
graphlayouts/man/graphlayouts.Rd 0000644 0001762 0000144 00000002514 13526724506 016574 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/graphlayouts.R
\docType{package}
\name{graphlayouts}
\alias{graphlayouts}
\alias{graphlayouts-package}
\title{graphlayouts: layout algorithms for network visualizations}
\description{
The package implements several new layout algorithms to visualize networks.
Most are based on the concept of stress majorization. Some more specific algorithms allow to emphasize
hidden group structures in networks or focus on specific nodes. The package is best used in conjunction with
ggraph.
}
\details{
Some features of the package are:
\itemize{
\item \code{layout_with_stress()} is a state of the art deterministic layout algorithms.
\item \code{layout_as_backbone()} uncovers hidden group structures (if they exist) by emphasizing strongly embedded edges.
\item \code{layout_with_focus()} and \code{layout_with_centrality()} produce concentric layouts with a focal or most central nodes in the center.
\item \code{layout_with_eigen()} implements some layout algorithms on the basis of eigenvectors
\item \code{layout_with_sparse_stress()} sparse stress for large graphs
\item \code{layout_with_pmds()} pivot MDS for large graphs.
\item \code{layout_as_dynamic()} for longitudinal network data
}
A detailed tutorial can be found \href{http://mr.schochastics.net/netVizR.html}{here}.
}
graphlayouts/man/layout_manipulate.Rd 0000644 0001762 0000144 00000001500 13526724506 017600 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/layout_manipulate.R
\name{layout_manipulate}
\alias{layout_manipulate}
\alias{layout_rotate}
\alias{layout_mirror}
\title{manipulate layout}
\usage{
layout_rotate(xy, angle)
layout_mirror(xy, axis = "vertical")
}
\arguments{
\item{xy}{graph layout}
\item{angle}{angle for rotation}
\item{axis}{mirror horizontal or vertical}
}
\value{
manipulated matrix of xy coordinates
}
\description{
functions to manipulate an existing layout
}
\details{
These functions are mostly useful for deterministic layouts such as \link{layout_with_stress}
}
\examples{
library(igraph)
g <- sample_gnp(50,0.3)
xy <- layout_with_stress(g)
#rotate 90 degrees
xy <- layout_rotate(xy,90)
# flip horizontally
xy <- layout_mirror(xy,"horizontal")
}
\author{
David Schoch
}
graphlayouts/man/layout_pmds.Rd 0000644 0001762 0000144 00000003156 13526725334 016415 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/layout_large_graphs.R, R/layouts.R
\name{layout_pmds}
\alias{layout_pmds}
\alias{layout_with_pmds}
\alias{layout_igraph_pmds}
\title{pivot MDS graph layout}
\usage{
layout_with_pmds(g, pivots, weights = NA)
layout_igraph_pmds(g, pivots, weights = NA, circular)
}
\arguments{
\item{g}{igraph object}
\item{pivots}{number of pivots}
\item{weights}{possibly a numeric vector with edge weights. If this is NULL and the graph has a weight edge attribute, then the attribute is used. If this is NA then no weights are used (even if the graph has a weight attribute). By default, weights are ignored. See details for more.}
\item{circular}{not used}
}
\value{
matrix of xy coordinates
}
\description{
similar to \link[igraph]{layout_with_mds} but uses only a small set of pivots for MDS. Considerably faster than MDS and thus applicable for larger graphs.
}
\details{
Be careful when using weights. In most cases, the inverse of the edge weights should be used to ensure that the endpoints of an edges with higher weights are closer together (weights=1/E(g)$weight)
The layout_igraph_* function should not be used directly. It is only used as an argument for plotting with 'igraph'.
'ggraph' natively supports the layout.
}
\examples{
\dontrun{
library(igraph)
library(ggraph)
g <- sample_gnp(1000,0.01)
xy <- layout_with_pmds(g,pivots = 100)
}
}
\references{
Brandes, U. and Pich, C. (2006). Eigensolver Methods for Progressive Multidimensional Scaling of Large Data. In \emph{International Symposium on Graph Drawing} (pp. 42-53). Springer
}
\author{
David Schoch
}
graphlayouts/man/layout_centrality.Rd 0000644 0001762 0000144 00000004131 13526724506 017622 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/layout_stress.R, R/layouts.R
\name{layout_centrality}
\alias{layout_centrality}
\alias{layout_with_centrality}
\alias{layout_igraph_centrality}
\title{radial centrality layout}
\usage{
layout_with_centrality(g, cent, scale = TRUE, iter = 500,
tol = 1e-04, tseq = seq(0, 1, 0.2))
layout_igraph_centrality(g, cent, scale = TRUE, iter = 500,
tol = 1e-04, tseq = seq(0, 1, 0.2), circular)
}
\arguments{
\item{g}{igraph object}
\item{cent}{centrality scores}
\item{scale}{logical. should centrality scores be scaled to \eqn{[0,100]}? (Default: TRUE)}
\item{iter}{number of iterations during stress optimization}
\item{tol}{stopping criterion for stress optimization}
\item{tseq}{numeric vector. increasing sequence of coefficients to combine regular stress and constraint stress. See details.}
\item{circular}{not used}
}
\value{
matrix of xy coordinates
}
\description{
arranges nodes in concentric circles according to a centrality index.
}
\details{
The function optimizes a convex combination of regular stress and a constrained stress function which forces
nodes to be arranged on concentric circles. The vector \code{tseq} is the sequence of parameters used for the convex combination.
In iteration i of the algorithm \eqn{tseq[i]} is used to combine regular and constraint stress as \eqn{(1-tseq[i])*stress_{regular}+tseq[i]*stress_{constraint}}. The sequence must be increasing, start at zero and end at one. The default setting should be a good choice for most graphs.
The layout_igraph_* function should not be used directly. It is only used as an argument for plotting with 'igraph'.
'ggraph' natively supports the layout.
}
\examples{
library(igraph)
library(ggraph)
g <- sample_gnp(10,0.4)
\dontrun{
ggraph(g,layout="centrality",centrality = closeness(g))+
draw_circle(use = "cent")+
geom_edge_link0()+
geom_node_point(shape = 21,fill = "grey25",size = 5)+
theme_graph()+
coord_fixed()
}
}
\references{
Brandes, U., & Pich, C. (2011). More flexible radial layout. Journal of Graph Algorithms and Applications, 15(1), 157-173.
}
graphlayouts/man/layout_stress.Rd 0000644 0001762 0000144 00000004071 13526724506 016772 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/layout_stress.R, R/layouts.R
\name{layout_stress}
\alias{layout_stress}
\alias{layout_with_stress}
\alias{layout_igraph_stress}
\title{stress majorization graph layout}
\usage{
layout_with_stress(g, weights = NA, iter = 500, tol = 1e-04,
mds = TRUE, bbox = 30)
layout_igraph_stress(g, weights = NA, iter = 500, tol = 1e-04,
mds = TRUE, bbox = 30, circular)
}
\arguments{
\item{g}{igraph object}
\item{weights}{possibly a numeric vector with edge weights. If this is NULL and the graph has a weight edge attribute, then the attribute is used. If this is NA then no weights are used (even if the graph has a weight attribute). By default, weights are ignored. See details for more.}
\item{iter}{number of iterations during stress optimization}
\item{tol}{stopping criterion for stress optimization}
\item{mds}{should an MDS layout be used as initial layout (default: TRUE)}
\item{bbox}{constrain dimension of output. Only relevant to determine the placement of disconnected graphs}
\item{circular}{not used}
}
\value{
matrix of xy coordinates
}
\description{
force-directed graph layout based on stress majorization.
}
\details{
Be careful when using weights. In most cases, the inverse of the edge weights should be used to ensure that the endpoints of an edges with higher weights are closer together (weights=1/E(g)$weight).
The layout_igraph_* function should not be used directly. It is only used as an argument for plotting with 'igraph'.
'ggraph' natively supports the layout.
}
\examples{
library(igraph)
library(ggraph)
set.seed(665)
g <- sample_pa(100,1,1,directed = FALSE)
# calculate layout manualy
xy <- layout_with_stress(g)
# use it with ggraph
\dontrun{
ggraph(g,layout = "stress")+
geom_edge_link0(edge_width = 0.2,colour = "grey")+
geom_node_point(col = "black",size = 0.3)+
theme_graph()
}
}
\references{
Gansner, E. R., Koren, Y., & North, S. (2004). Graph drawing by stress majorization. \emph{In International Symposium on Graph Drawing} (pp. 239-250). Springer, Berlin, Heidelberg.
}
graphlayouts/man/layout_focus.Rd 0000644 0001762 0000144 00000003277 13526724506 016575 0 ustar ligges users % Generated by roxygen2: do not edit by hand
% Please edit documentation in R/layout_stress.R, R/layouts.R
\name{layout_focus}
\alias{layout_focus}
\alias{layout_with_focus}
\alias{layout_igraph_focus}
\title{radial focus layout}
\usage{
layout_with_focus(g, v, weights = NA, iter = 500, tol = 1e-04)
layout_igraph_focus(g, v, weights = NA, iter = 500, tol = 1e-04,
circular)
}
\arguments{
\item{g}{igraph object}
\item{v}{id of focal node to be placed in the center}
\item{weights}{possibly a numeric vector with edge weights. If this is NULL and the graph has a weight edge attribute, then the attribute is used. If this is NA then no weights are used (even if the graph has a weight attribute). By default, weights are ignored. See details for more.}
\item{iter}{number of iterations during stress optimization}
\item{tol}{stopping criterion for stress optimization}
\item{circular}{not used}
}
\value{
a list containing xy coordinates and the distances to the focal node
}
\description{
arrange nodes in concentric circles around a focal node according to their distance from the focus.
}
\details{
Be careful when using weights. In most cases, the inverse of the edge weights should be used to ensure that the endpoints of an edges with higher weights are closer together (weights=1/E(g)$weight).
The layout_igraph_* function should not be used directly. It is only used as an argument for plotting with 'igraph'.
'ggraph' natively supports the layout.
}
\examples{
library(igraph)
library(ggraph)
g <- sample_gnp(10,0.4)
coords <- layout_with_focus(g,v = 1)
coords
}
\references{
Brandes, U., & Pich, C. (2011). More flexible radial layout. \emph{Journal of Graph Algorithms and Applications}, 15(1), 157-173.
}
graphlayouts/man/figures/ 0000755 0001762 0000144 00000000000 13526724506 015225 5 ustar ligges users graphlayouts/man/figures/README-backbone-1.png 0000644 0001762 0000144 00000552355 13522271455 020603 0 ustar ligges users PNG
IHDR Hc pHYs od IDATxw| {WeْeYnd lpc @!1p $$ !9'^LuŽJ%wu/W0XIx |D"!JgΜH$
,,j_z%h D"q8 JRP\㝴`=` hv:fj5B+JsssCVeYvdd$L^``0̗ ÔfM600DB6
^DQJ d <`Xldd(\nx< B(N#B!CDeesb,1tt:z~:I x \&!ɤӝ`r B$IX,
@ ÜkZXvy_gggEE`\56! .
Ln{^,dG"$ p-HgH2f,4ެqG|^2k6q 2KU8 v `r6Wf}
e
.-P~?KJJD"Q0O x/oQX,2XG#H;::ÂrabP%*п2z^jX|>Fq\bz"sX,^ D< P(loocY^RY־7#/nj2eeer֜$iSwzȷ}/Q{&4{3]qnCh4ZSS/_Rdnn
{2s-PH&M' #sa 8 |WBӟN ӧOyyy+W+ |0so'O۷oÆO !$ֿ ɤ`** E@ _Td2hAQT*ť OϚ5kdd
[F#EQ^4Zң#~Р 8s!b܅?k?*
<~ v
h47|
꾉5% P$E}LN ]Y+ \@EQ0;VM:A(Jy^l IDAT< _5MAA%Q
fAG*A, =@!7@윜 !fKʎ&
B vB*:w\IIIaa!˲}}}@ \."a4
'_'I mp,@6 !X- )%mmm]]]k˻[?cVFK4 mc"czvRF\ 5)}SO> ]bŊ믿~dd+?? (F_߿ƍч} |NҐ'xhB co1iFs&sR={ްaŒWo@fXJ|>@ ڕJfQUQQH$W<
oh?8 Bw F}~\,+6!e].B\_:,([jPǥ8qT{Z .2֚AW:G&B-9 {aB=N0 F Yn|H$Ir1/r e%I ໃ;!460 elG#4D"& -J'&&}h4<66F^Tn#r5۾IRDD"xyx<Au
.\غuZ!rNja fyȑ;ＳV
Cb9 NOO jŃ0 aMo!z[VF\5^.)T^*mhaaeevM
U*{U*C=fvzÀ`{"5*:WFCf24QͦH$2 XL,_|yxxX,zrZz GRAfTuҥx
,Aq- f.ۇBt:P( C F# `~8 >Ξ=[5M&yw qKGFDb!^l?vK/7EQFvLvsn'$G#YXXD" Jp8RTӉt;wn;77'j56Zx5l6{< x s˙L-[ e.ݻ&"+ѩJyf``qPPPVV O>H$Yǎ ߈vK\5s^P(Æbnݪ0$=@GjرEEEѦG ]bQpp0oV.ie 莘,YБrU%b!_zenn+Rp]\\rrr_ncc3uT[[ŋsfq.%%%|OZ
!sNðI& qt(,,400;DU]]DI۷o՞ d @ ~~~#~ @OOٳgݻGw⛚o߮S[ srU4ͪQUUCBBTׯ_͛pVڵk ;J?ëW` w%Kl߾ (,R1\jkdd4g +++ધ'11(!!^ϝ ']F/* APBW3
*(3QO>e(L g8Ir
yxxpnd~15S fll[*jll,$$$'+&&aVTϟ?߿?666 `+ IDATʕx<^YYYCCCNNȑ#W\)))'z2F;R|r^n @(//L3VPV) H:{/22B^~u֭zzzEEE~~~vvvpݍF-ZzϽom6&S
HLLdRruuOgtn} .===NP
H$k =A
+LP x[>?F=wKVC {zz4g``ĉHHgysh,3gμtҶm۬-,,6nܘRSS=ܹ-r|Ad<<} !loo{.{yy͚5缃A&/^xHi4
Ú`: 2<#Ś1cFJJ
2Ȇ0 Hj؈1c
J
_/hw?h2tmT E̙[vD$nS'%KqgbСCb z^t24ͣ y'98tn:0h۫*333*
}OQҨ2gD\1ů)l\f(ctvvDP(\.fv$ P(8 &{`PNblq6+w֢Q*I(Y`h
m'5 8 {Ny&ؼżaПS;|o?Td
W*#ܞDrV&u,kpp0Hy<[[OZŖ~rd4=vعsơCFGGקr\^ڲ_'bF$J
B,kkk