V. Batagelj Introduction to Network Analysis using...
Transcript of V. Batagelj Introduction to Network Analysis using...
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Introduction to Network Analysis using Pajek
4. Structure of networks 2Connectivity
Vladimir Batagelj
IMFM Ljubljana and IAM Koper
Phd program on StatisticsUniversity of Ljubljana, 2018
1 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Outline
1 Connectivity2 Condensation3 Bow-tie4 Other connectivities5 Important nodes6 Closeness7 Betweeness8 Hubs and authorities9 Clustering
4
4
4
32
2
3
3
3
2
456
5
56
3 2
BCD A
1
1
Self colours
Title
Quad Royal Tube Map Version 2 Date
Size
1/9/2005
4 Colour Process + 4 Self Colours S/S
Pantone®
485 CPantone®
470 CPantone®
235 CPantone®
072 C
Process colours
River Thames
River Thames
A
B
C
D
E
F
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
A
B
C
D
E
F
Tubemap
D2 Acton CentralD2 Acton TownD6 AldgateD7 Aldgate EastD8 All SaintsB2 AlpertonA1 AmershamC6 AngelB5 ArchwayA6 Arnos GroveB6 Arsenal
C4 Baker Street F4 BalhamD6 BankC6 BarbicanC9 BarkingB9 BarkingsideD3 Barons CourtC3 BayswaterE9 Beckton
D9 Beckton ParkC9 BecontreeB5 Belsize ParkD6 BermondseyC7 Bethnal GreenD5 BlackfriarsB7 Blackhorse RoadD8 BlackwallC4 Bond StreetE6 BoroughD1 Boston ManorA6 Bounds GreenC8 Bow ChurchC7 Bow RoadB4 Brent CrossF5 Brixton C8 Bromley-by-BowB3 BrondesburyB3 Brondesbury ParkA8 Buckhurst Hill A4 Burnt Oak
B6 Caledonian RoadB6 Caledonian Road
& BarnsburyB5 Camden RoadB5 Camden TownD7 Canada WaterD8 Canary WharfD8 Canning TownD6 Cannon StreetB7 CanonburyA3 Canons ParkA1 Chalfont & LatimerB5 Chalk FarmC5 Chancery LaneD5 Charing CrossA1 CheshamA9 ChigwellD2 Chiswick ParkA1 ChorleywoodF4 Clapham CommonF4 Clapham North F4 Clapham SouthA6 Cockfosters
A4 ColindaleF4 Colliers WoodD5 Covent GardenE8 Crossharbour
& London ArenaA2 CroxleyD9 Custom HouseF8 Cutty SarkD9 Cyprus
B9 Dagenham EastB9 Dagenham HeathwayB7 Dalston KingslandA8 DebdenF7 Deptford BridgeC8 Devons RoadB3 Dollis Hill
C1 Ealing BroadwayD2 Ealing CommonD3 Earl's CourtC2 East ActonA2 EastcoteA5 East Finchley
C8 East HamD8 East IndiaE3 East PutneyA4 EdgwareC4 Edgware Road (Bakerloo)C4 Edgware Road
(Circle/District/H&C)E5 Elephant & CastleB9 Elm ParkF7 Elverson RoadD5 EmbankmentA8 EppingC5 EustonC5 Euston Square
B9 FairlopC6 FarringdonA5 Finchley CentralB4 Finchley RoadB4 Finchley Road & FrognalB6 Finsbury ParkE3 Fulham Broadway
E9 Gallions Reach
B8 Gants HillD3 Gloucester RoadB4 Golders GreenD3 Goldhawk RoadC5 Goodge StreetB5 Gospel OakA9 Grange Hill C5 Great Portland StreetB1 GreenfordF7 GreenwichD4 Green ParkE2 Gunnersbury
B7 Hackney CentralB7 Hackney WickA9 HainaultD3 HammersmithB5 HampsteadB5 Hampstead HeathC2 Hanger LaneB3 HarlesdenA3 Harrow & Wealdstone B2 Harrow-on-the HillE1 Hatton Cross
E1 Heathrow Terminals 1, 2, 3
E1 Heathrow Terminal 4A4 Hendon CentralD8 Heron QuaysA5 High BarnetB6 Highbury & IslingtonA5 HighgateD3 High Street KensingtonA1 HillingdonC5 Holborn C3 Holland ParkB6 Holloway RoadB7 HomertonB9 HornchurchE1 Hounslow CentralD1 Hounslow EastE1 Hounslow WestD4 Hyde Park Corner
A1 IckenhamE8 Island Gardens
E5 Kennington
B3 Kensal GreenB3 Kensal RiseD3 Kensington (Olympia)B5 Kentish TownB5 Kentish Town WestA3 KentonE2 Kew GardensB4 KilburnC3 Kilburn ParkB3 KingsburyC5 King’s Cross St. PancrasD4 Knightsbridge
C3 Ladbroke GroveE5 Lambeth NorthC4 Lancaster GateC3 Latimer RoadD5 Leicester SquareF7 LewishamB8 LeytonB8 LeytonstoneD7 LimehouseC6 Liverpool StreetD6 London Bridge
A8 Loughton
C3 Maida ValeB6 Manor HouseD5 Mansion HouseC4 Marble ArchC4 MaryleboneC7 Mile EndA5 Mill Hill EastD6 MonumentC6 MoorgateA2 Moor ParkF4 MordenB5 Mornington Crescent E8 Mudchute
B3 NeasdenB9 Newbury ParkF7 New CrossF7 New Cross GateC2 North ActonC2 North EalingD1 NorthfieldsD8 North Greenwich
A2 North HarrowB1 NortholtB3 North WembleyB3 Northwick ParkA2 NorthwoodA2 Northwood HillsE9 North WoolwichC3 Notting Hill Gate
A6 OakwoodC6 Old StreetD3 OlympiaD1 OsterleyF5 Oval C4 Oxford Circus
C3 PaddingtonC2 Park RoyalE3 Parsons GreenC1 PerivaleD5 Piccadilly CircusE4 PimlicoA2 PinnerC8 Plaistow
D8 PoplarB3 Preston RoadD9 Prince RegentC8 Pudding Mill LaneE3 Putney Bridge
A3 QueensburyB3 Queen’s ParkC3 Queensway
D3 Ravenscourt ParkB2 Rayners LaneB8 RedbridgeC4 Regent’s ParkE2 RichmondA1 RickmansworthA8 Roding ValleyD7 RotherhitheD9 Royal AlbertC3 Royal OakD9 Royal VictoriaA1 RuislipB1 Ruislip GardensA2 Ruislip Manor
C5 Russell Square
D4 St. James’s ParkC4 St. John’s WoodC6 St. Paul’sB7 Seven SistersD7 ShadwellC3 Shepherd’s Bush
(Central)D3 Shepherd’s Bush
(Hammersmith & City)C7 Shoreditch E9 SilvertownD4 Sloane SquareB8 SnaresbrookD2 South ActonD2 South EalingE3 SouthfieldsA6 SouthgateB2 South HarrowD4 South KensingtonB3 South KentonE8 South QuayB1 South Ruislip
E5 SouthwarkF4 South WimbledonB8 South WoodfordD2 Stamford BrookA3 StanmoreC7 Stepney GreenF5 StockwellB3 Stonebridge ParkC8 StratfordB2 Sudbury HillB2 Sudbury Town E7 Surrey QuaysB4 Swiss Cottage
D5 TempleA8 Theydon BoisF4 Tooting BecF4 Tooting BroadwayC5 Tottenham
Court RoadB7 Tottenham HaleA5 Totteridge &
WhetstoneD7 Tower Gateway
D6 Tower HillB5 Tufnell ParkD2 Turnham Green A6 Turnpike Lane
B9 UpminsterB9 Upminster BridgeC9 UpneyC8 Upton ParkA1 Uxbridge
E4 VauxhallD4 Victoria
B7 Walthamstow CentralB8 WansteadD7 WappingC5 Warren StreetC3 Warwick AvenueE5 WaterlooA2 WatfordB3 Wembley CentralB3 Wembley ParkC2 West Acton
C3 Westbourne ParkD3 West BromptonD7 WestferryA5 West FinchleyC8 West HamB4 West HampsteadB2 West HarrowD8 West India QuayD3 West KensingtonD5 WestminsterA1 West RuislipC7 WhitechapelC3 White CityB3 Willesden GreenB3 Willesden JunctionE3 WimbledonE3 Wimbledon ParkA8 WoodfordA6 Wood GreenA5 Woodside Park
µ:
Ÿ Á :
Ÿ Á
:
Á:
Á : µŸ :
Á µ
Á µÁ :
Á µ
Ÿ ÁÁ µ
:
ÁµÁŸ ÁÁ
Ÿ Á :Ÿ
µ
Á : µ: µÁ : µ
Ÿ Á :Ÿ Á : µ
:
Ÿ Á : µ
ÁŸ Á : µÁÁÁŸ Á :
Ÿ Á
µ
Ÿ Á :Á µÁ µÁ µ
Á: µ
Ÿ ÁÁ µµ:
Á :
µ
Ÿ :Ÿ Á
Á µÁ µÁŸ Á
µÁ µ
Ÿ Á : µ: ∏
Ÿ ::
Ÿ Á
Á :µ
Á µ
Á :Á
:
Ÿ Á :Á µ:
Á
Ÿ Á :Á µ
Á :
Ÿ µŸ Á :Ÿ Á
: µ ∏
:
µŸ Á :ÁŸ
Ÿ Á µ
Á
Ÿ ÁÁŸ : µŸ Á µ
Ÿ Á: µ
Á
: µ
Á : µÁ : µ Á :Á : ∏
ÁÁ
Á : µ:
Ÿ :µÁ : ∏: µ
Ÿ Á
Á :
Ÿ
Ÿ Á :Ÿ :
µ
:
Ÿ Á :Á µ
Ÿ Á :Á :Ÿ Á : µ
Á :Á :
Á :Ÿ Á ::
Ÿ Á :
: µŸ Á :
: :
ÁŸ Á :: ∏
Ÿ Á :Ÿ Á
Á : µ:
Á µµÁ
Ÿ Á :
ÁŸ Á :Ÿ Á :
Ÿ Á : µŸ Á :
ÁÁ µ
Á µŸ Á :Ÿ ::
Á :
Á :
Ÿ Á :
Ÿ:
µŸ Á
Á µ
Ÿ Á :Á :Ÿ Á : µ
µ:
Ÿ Á : µÁÁ
Ÿ :
Ÿ Á : µŸ Á :
µ
Ÿ Á : µ
: µ
Á µ
: ∏
Ÿ ÁŸ
Á :Ÿ Á ::
Ÿ Á :Á :
µ: µµ:
:
µ
µŸ ::
:
Á :Á : µŸ Á : µ
Ÿ Á : µÁŸ Á : µ
231124D1
2/342
131145213
35222132114
3/42232
2/32254
22
22223125C211D53B22
2/35
4312
A3
2/33
5526
2/323
33
1/2253
3/42/32/3511
1/26
2/31611
5142222
3
413212514
2/313
2252
2/333355
5/6
6
63/4252316122264451
62
2
222224
3/422411
21121
2/33
3/4211
6
22/31112411
6/A422
3422
2/333
2/3
5544663
1/2
512421
13241153
243
2/32
421
25414A52323656
1
121322
2
2314333451425
13/442522334422
16331
34
1
12
2/33
66436
1/21
342121A443
2224325221622
2/3333434
A
B
C
D
F
G
H
I
K
L
M
N
O
Q
P
R
S
T
V
W
U
E
Grid Stations Zones Grid Stations Facilities Zones
Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities Zones Grid Stations Facilities ZonesIndex to stations
Sponsored by
Travel information
Step-free accessStations displaying this symbolin the index have step-freeaccess between the street andplatforms. This facility is usefulfor passengers with luggage,shopping or buggies as well asfor wheelchair users
To plan a journey in a wheelchair,see our leaflet ‘Tube access guide’ or call
0845 330 9880
For journey planning and travel advice call
µ
Station facilitiesThe index on this map also shows
Other RailwaysFor a map of all Railways in Greater London, consult the High Frequency Services Map nearby
‰ Car parksBicycle parking
Stations with toilets on siteor nearby
Á
∑
Travel Information Centres∏
Facilities
Transport for London
Key to lines and symbols
Central
Bakerloo
District
Circle
East London Docklands Light Railway
Northern
Metropolitan
Piccadilly
Victoria
Waterloo & City
National Rail
Connections withNational Rail
Connection withTramlink
Airport interchange
Connections withriverboat services
Interchange stations
Poster 09.05
OpensDecember 2005
020 7222 123424 hour travel information
020 7918 3015Textphone
www.tfl.gov.ukWebsite
020 7918 3015Textphone
www.tfl.gov.ukWebsite
Jubilee
Hammersmith & City
Covent Garden station gets very busy at weekends and in the evenings, but you can avoid the crowds by walking there from Holborn, Leicester Square or Charing Cross. The short walk is clearly signposted above ground and maps are on display at each station.
This diagram is an evolution of the original design conceived in 1931 by Harry Beck
Bermondsey
SouthwarkWaterloo East
Chalfont &Latimer
Moor Park
NorthwoodNorthwoodHills
Pinner
Eastcote North Harrow
Maida Vale
Queen's ParkKensal Green
Neasden
Dollis Hill
Willesden Green
Kilburn
WestHampstead
Swiss CottageSt. John's Wood
Finchley Road
Amersham
Ruislip Manor
Chesham
Chorleywood
Rickmansworth
Watford
Croxley
Harrow-on-the-Hill
PrestonRoad
Hillingdon Ruislip
Rayners Lane
West Harrow NorthwickPark Wembley
Park
Ealing Common
EalingBroadway
GreatPortland
StreetBakerStreet
FarringdonBarbican
Moorgate
Aldgate
EustonSquare
ActonTown
ChiswickPark
TurnhamGreen
WestActon
EastActon
Shepherd'sBush
StamfordBrook
RavenscourtPark
Hammersmith
WestKensington
West Brompton
Fulham Broadway
Parsons Green
Putney Bridge
East Putney
Southfields
Wimbledon Park
Wimbledon
VictoriaSouthKensington
GloucesterRoad
Embankment
Blackfriars
MansionHouse
Temple
Cannon Street
Bank
Monument
BaronsCourt
Fenchurch Street
Whitechapel
TowerGateway
TowerHill
AldgateEast
Stepney Green
Mile End
BowRoad Bow
ChurchBromley-by-Bow
West Ham
Plaistow
Upton Park
East Ham
Becontree
DagenhamHeathway
Elm Park
Upney
DagenhamEast
Hornchurch
UpminsterBridge
Upminster
High StreetKensington
NottingHill Gate
Bayswater
Kensal Rise Brondesbury
EdgwareRoad
St. James'sPark
SloaneSquare
Westminster
Barking
Latimer Road
Westbourne Park
Finchley Road& Frognal
Ladbroke Grove
Royal Oak
Shepherd'sBush
Goldhawk Road
West Ruislip
Greenford
RuislipGardens
SouthRuislip
Northolt
HangerLane
Perivale
NorthActon
WhiteCity
HollandPark
Paddington
Paddington
ChanceryLaneBond
StreetOxfordCircus
TottenhamCourt Road
St. Paul'sMarbleArch
Queensway
LancasterGate
BethnalGreen
Stratford
Leyton
Leytonstone
Snaresbrook
SouthWoodford
Woodford
Epping
Theydon Bois
DebdenLoughton
Buckhurst Hill
Redbridge
ChigwellRodingValley
Hainault
Fairlop
BarkingsideNewbury
Park
GrangeHill
Wanstead GantsHill
South Ealing
Knightsbridge
Hyde ParkCorner
Green Park
PiccadillyCircus
LeicesterSquare
RussellSquare
Caledonian Road
CaledonianRoad &
Barnsbury
DalstonKingsland
Homerton
Holloway Road
Arsenal
Manor House
Turnpike Lane
Wood Green
Bounds Green
Arnos Grove
Southgate
Oakwood
Cockfosters
Uxbridge Ickenham
ActonCentral
Waterloo
Morden
Colliers Wood Tooting Broadway
South Wimbledon
Tooting BecBalham
Clapham South ClaphamCommon
Clapham NorthClapham High Street 100m
StockwellOval
Kennington
Borough
SouthActon
Old Street
Angel
GoodgeStreet
Euston
MorningtonCrescent
Camden Town
Chalk Farm
Regent’s Park
Belsize Park
Hampstead HampsteadHeath
GospelOak
CanonburyHackneyCentral
HackneyWick
KentishTown West
CamdenRoad
Hendon Central
Colindale
BurntOak
Mill Hill East
High Barnet
Totteridge & Whetstone
Woodside Park
West Finchley
Finchley Central
East Finchley
Highgate
Archway
Tufnell Park
KentishTown
CanadaWater
Canary Wharf
Elverson Road
Deptford Bridge
Harrow &Wealdstone
Kenton
Stanmore
Canons Park
Queensbury
Kingsbury
South KentonNorth Wembley
Wembley Central
Stonebridge ParkHarlesden
Willesden Junction
Kilburn ParkWarwick Avenue
EdgwareRoad
BrondesburyPark
Marylebone
LambethNorth
Elephant & Castle
King's CrossSt. Pancras
CharingCross
Covent Garden
Highbury &Islington
BlackhorseRoad
SevenSisters
WalthamstowCentral
TottenhamHale
FinsburyPark
Pimlico
Brixton
Shoreditch
Wapping
Rotherhithe
Surrey Quays
New CrossNew Cross Gate
Vauxhall
Limehouse
Westferry
DevonsRoad
PuddingMill Lane
West IndiaQuay
Cutty Sarkfor Maritime Greenwich
Greenwich
Lewisham
Blackwall
EastIndia
Warren Street
Edgware
All Saints
Heron Quays
South Quay
Crossharbour &London Arena
Mudchute
Island Gardens
Shadwell
No service between Woodford - Hainault
after 2000 hours
Gunnersbury
Richmond
Kew Gardens
Poplar
London Bridge
Change atChalfont & Latimer
on most trains
No Piccadilly line serviceUxbridge - Rayners Lane
in the early mornings
Special fares apply forprinted single and return
tickets to and from this station
Also served byPiccadilly linetrains early
mornings andlate evenings
100m
100m
Euston 200m
150m
Charing Cross 100m
200m
HeathrowTerminals
1, 2, 3
Hounslow Central
Osterley
Northfields
Boston ManorHounslow
East
HounslowWest
LiverpoolStreet
No Hammersmith & City line serviceWhitechapel - Barking early mornings,late evenings or all day Sundays.
Mondays - Fridays open0700 - 1030 and 1530 - 2030
Saturdays closedSundays open 0700 - 1500
No entry from the streeton Sundays 1300 - 1730
(exit and interchange only)
Waterloo & City lineMondays - Fridays 0615 - 2130
Saturdays 0800 - 1830Sundays closed
At off-peak times most trains run to/from Morden via the Bank branch.To travel to/from the Charing Cross branch please change at Kennington.
Open Mondays -Saturdays
200m
South Harrow
Sudbury Hill
North Ealing
Park Royal
Alperton
Sudbury Town
Mondays - Saturdaysopen 0700-2345
Sundaysopen 0800-2345
Kensington(Olympia)
Earl'sCourt
Holborn
King George V
LondonCityAirport
WestSilvertown
PontoonDock
Opens
December 2005
Opens December 2005
Silvertown
North Woolwich
Royal Victoria
Custom Housefor ExCeL
Prince Regent
Royal Albert
Beckton Park
Cyprus
GallionsReach
Beckton
Canning TownBus to London City Airport
OpenMondays - Fridays
until 2100 onlySaturdays 0730 - 1930
Golders Green
Brent Cross
HeathrowTerminal 4
Hatton Crossfor Heathrow Terminal 4
Bus ser
vice
Improvement work to tracks and stations mayaffect your journey, particularly at weekends.
For help planning your journey look forpublicity at stations, call 020 7222 1234
or visit www.tfl.gov.uk
Closed until May 2006
Sudbury Hill Harrow150m
Station in Zone 11
Station in Zone 2
3
4
5
6
Station in Zone 3
Station in Zone 5
Station in Zone 6
2
Station in Zone 4
A Station in Zone A
B Station in Zone B
C Station in Zone C
D Station in Zone D
Station in Zone 6 and Zone A
Station in both zones
Station in both zones
Explanation of zones
TM Quad 2n Version 2 1/9/05 4 Colour Process + 4 Self Colours
NorthGreenwich
e-mail: [email protected] version of slides (April 17, 2018 at 02 : 59):http://vladowiki.fmf.uni-lj.si/doku.php?id=pajek:ev:pde
2 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Walks
length |s| of the walk s is the num-ber of links it contains.s = (j , h, l , g , e, f , h, l , e, c , b, a)|s| = 11A walk is closed iff its initial andterminal node coincide.If we don’t consider the directionof the links in the walk we get asemiwalk or chain.trail – walk with all links differentpath – walk with all nodes differ-entcycle – closed walk with all internalnodes different
A graph is acyclic if it doesn’t contain any cycle.
3 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Shortest paths
A shortest path from u to v isalso called a geodesic from uto v . Its length is denoted byd(u, v).If there is no walk from u to vthen d(u, v) =∞.d(j , a) = |(j , h, d , c , b, a)| = 5d(a, j) =∞d(u, v) = max(d(u, v), d(v , u))is a distance:d(v , v) = 0, d(u, v) = d(v , u),d(u, v) ≤ d(u, t) + d(t, v).
The diameter of a graph equals to the distance between the mostdistant pair of nodes: D = maxu,v∈V d(u, v).
Network/Create New Network/Subnetwork with Paths/
4 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Shortest paths
black
backlack clackblank
rack wackbalk bank basklick lanklace clickclankblink
rick race wickwalk bilkbale bane bastlinklice lanelate clink chick
rice rate winkwale bilehale wanebine wastbaitline chinkcline chic
rite winewilewhale wait chine chit
write whinewhile whit
white
DICT28.
5 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Equivalence relations and Partitions
A relation R on V is an equivalence relation iff it isreflexive ∀v ∈ V : vRv , symmetric ∀u, v ∈ V : uRv ⇒ vRu, andtransitive ∀u, v , z ∈ V : uRz ∧ zRv ⇒ uRv .
Each equivalence relation determines a partition into equivalenceclasses [v ] = {u : vRu}.
Each partition C determines an equivalence relationuRv ⇔ ∃C ∈ C : u ∈ C ∧ v ∈ C .
k-neighbors of v is the set of nodes on ’distance’ k from v ,Nk(v) = {u ∈ v : d(v , u) = k}.The set of all k-neighbors, k = 0, 1, ... of v is a partition of V.
k-neighborhood of v , N(k)(v) = {u ∈ v : d(v , u) ≤ k}.
Network/Create Partition/k-Neighbors
6 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Motorola’s neighborhood
####
AOL
MSFT
Yahoo!
Sun
IBM
Cisco
iPlanet
3COM
Motorola
AvantGo
BaltimoreTechnologies
Unisys
ComputerAssociates
Nextel
MapInfo
GM
MasterCard
Fujitsu
55
1
365
40
46
1
33
19
23
4
22
2
1
4
6
1
1 1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
11
1
1
1
1
1
1
1
1
1
1
1
1
11
####
AOL
MSFT
Yahoo!
Sun
IBM
Cisco
iPlanet
3COM
Motorola
AvantGo
BaltimoreTechnologies
Unisys
ComputerAssociates
Nextel
MapInfo
GM
MasterCard
Fujitsu
The thickness of edges is a square root of its value.
7 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Connectivity
Node u is reachable from node v iffthere exists a walk with initial nodev and terminal node u.
Node v is weakly connected withnode u iff there exists a semiwalkwith v and u as its end-nodes.
Node v is strongly connected withnode u iff they are mutually reach-able.
Weak and strong connectivity are equivalence relations.Equivalence classes induce weak/strong components.
Network/Create Partition/Components/
8 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Weak components
Reordering the nodes of net-work such that the nodes fromthe same class of weak partitionare put together we get a ma-trix representation consisting ofdiagonal blocks – weak compo-nents.
Most problems can be solvedseparately on each componentand afterward these solutionscombined into final solution.
9 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Special graphs – bipartite, tree
A graph G = (V,L) is bipartite iff its set of nodes V can bepartitioned into two sets V1 and V2 such that every link from L hasone end-node in V1 and the other in V2.
A weakly connected graph G is a tree iff it doesn’t contain loops andsemicycles of length at least 3.
10 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Condensation
If we shrink every strong component of a given graph into a node,delete all loops and identify parallel arcs the obtained reduced graphis acyclic. For every acyclic graph an ordering / level functioni : V → N exists s.t. (u, v) ∈ A ⇒ i(u) < i(v).
11 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Condensation – Example
Network/Create Partition/Components/Strong [1]Operations/Network+Partition/Shrink Network [1][0]Network/Create New Network/Transform/Remove/Loops [yes]Network/Acyclic Network/Depth Partition/AcyclicPartition/Make PermutationPermutation/Inverse Permutationselect partition [Strong Components]Operations/Partition+Permutation/Functional Composition Partition*PermutationPartition/Make Permutationselect [original network]File/Network/Export Matrix to EPS/Using Permutation
a
b
c
d
e
f
g
h
i
jk
l
#a
#e
f g
i
k
Pajek - shadow [0.00,1.00]
i
e
h
j
l
a
b
c
d
g
f
k
i e h j l a b c d g f k
12 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Internal structure of strong components
Let d be the largest common divisor oflengths of closed walks in a strong com-ponent.
The component is said to be simple, iffd = 1; otherwise it is periodical with aperiod d .
The set of nodes V of strongly connecteddirected graph G = (V,R) can be parti-tioned into d clusters V1, V2, . . . , Vd , s.t.for every arc (u, v) ∈ R holds u ∈ Vi ⇒v ∈ V(imod d)+1 .
Network/Create Partition/
Components/Strong-Periodic
13 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
. . . Internal structure of strong components
Bonhoure’s periodical graph. Pajek data
14 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Bow-tie structure of the Web graph
Kumar &: The Web as a graph
Let S be the largest strong com-ponent in network N ;W the weakcomponent containing S; I theset of nodes from which S canbe reached; O the set of nodesreachable from S; T (tubes) setof nodes (not in S) on paths fromI to O; R =W \ (I ∪S ∪O∪T )(tendrils); and D = V \ W. Thepartition
{I,S,O, T ,R,D}
is called the bow-tie partition of V.
Note: chains can exist in the set R.
Network/Create Partition/Bow-Tie
15 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Biconnectivity
Nodes u and v are biconnected iff they are connected (in bothdirections) by two independent (no common internal node) paths.Biconnectivity determines a partition of the set of links.
A node is an articulation node iff its deletion increases the number ofweak components in a graph.
A link is a bridge iff its deletion increases the number of weakcomponents in a graph.
Network/Create New Network/with Bi-Connected
Components.
The notion of biconnectivity can be generalized do k-connectivity. Novery efficient algorithm for k > 3 exists.
16 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
k-connectivity
Node connectivity κ of graph G is equal to the smallest number ofnodes that, if deleted, induce a disconnected graph or the trivialgraph K1.
Link connectivity λ of graph G is equal to the smallest number oflinks that, if deleted, induce a disconnected graph or the trivial graphK1.
Whitney’s inequality: κ(G) ≤ λ(G) ≤ δ(G) .
Graph G is (node) k−connected, if κ(G) ≥ k and is linkk−connected, if λ(G) ≥ k .
Whitney / Menger theorem: Graph G is node/link k−connected iffevery pair of nodes can be connected with k node/link internallydisjoint (semi)walks.
17 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Triangular and short cycle connectivities
In an undirected graph we call a triangle a subgraph isomorphic toK3.
A sequence (T1,T2, . . . ,Ts) of triangles of G (node) triangularlyconnects vertices u, v ∈ V iff u ∈ T1 and v ∈ Ts or u ∈ Ts andv ∈ T1 and V(Ti−1) ∩ V(Ti ) 6= ∅, i = 2, . . . s. It edge triangularlyconnects vertices u, v ∈ V iff a stronger version of the secondcondition holds E(Ti−1) ∩ E(Ti ) 6= ∅, i = 2, . . . s.
Node triangular connectivity is an equivalence on V; and edgetriangular connectivity is an equivalence on E . See the paper.
18 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Triangular connectivity in directed graphs
If a graph G is mixed we replaceedges with pairs of opposite arcs.In the following let G = (V,A)be a simple directed graph with-out loops. For a selected arc(u, v) ∈ A there are only two dif-ferent types of directed triangles:cyclic and transitive.
cyc tra
For each type we get the corresponding triangular network Ncyc andNtra by determining the corresponding weight wcyc or wtra to its arcs,counting the number of cyclic/transitive triangles that contain thearc. We remove arcs with weight zero. The notion of triangularconnectivity can be extended to the notion of short (semi) cycleconnectivity.
Network/Create New Network/with Ring Counts/3-Rings/Directed
19 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Edge-cut at level 16 of triangular networkof Erdos collaboration graph
AJTAI, MIKLOS
ALAVI, YOUSEF
ALON, NOGA
ARONOV, BORIS
BABAI, LASZLO
BOLLOBAS, BELA
CHARTRAND, GARY
CHEN, GUANTAO
CHUNG, FAN RONG K.
COLBOURN, CHARLES J.FAUDREE, RALPH J.
FRANKL, PETER
FUREDI, ZOLTANGODDARD, WAYNE D.
GRAHAM, RONALD L.
GYARFAS, ANDRAS
HARARY, FRANK
HEDETNIEMI, STEPHEN T.
HENNING, MICHAEL A.
JACOBSON, MICHAEL S.
KLEITMAN, DANIEL J.
KOMLOS, JANOS
KUBICKI, GRZEGORZ
LASKAR, RENU C.
LEHEL, JENO
LINIAL, NATHAN
LOVASZ, LASZLO
MAGIDOR, MENACHEMMCKAY, BRENDAN D.
MULLIN, RONALD C.
NESETRIL, JAROSLAV
OELLERMANN, ORTRUD R.
PACH, JANOS
PHELPS, KEVIN T.
POLLACK, RICHARD M.
RODL, VOJTECHROSA, ALEXANDER
SAKS, MICHAEL E.
SCHELP, RICHARD H.
SCHWENK, ALLEN JOHN
SHELAH, SAHARON
SPENCER, JOEL H.
STINSON, DOUGLAS ROBERT
SZEMEREDI, ENDRE
TUZA, ZSOLT
WORMALD, NICHOLAS C.
without Erdos,n = 6926,m = 11343
20 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Arc-cut at level 11 of transitive triangular networkof ODLIS dictionary
abstract
American Library Association /ALA/
American Library Directory
bibliographic record
bibliography
binding
blanket order
book
book size
Books in Print /BIP/
call number
catalog
charge
collation
colophon
condition
copyright
cover
dummy
dust jacket
edition
editor
endpaper
entry
fiction
fixed location
folio
frequency
front matter
half-title
homepage
imprint
index
International Standard Book Number /ISBN/
invoice
issue
journal layout
librarian
library
library binding
Library Literature
new book
Oak Knoll
page
parts of a book
periodical
plate
printing
publication
published price
publisher
publishing
review
round table
serial
series
suggestion box
table of contents /TOC/text
title
title page
transaction log
vendor
work
Pajek
21 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Important vertices in network
To identify important / interesting elements (nodes, links) in anetwork we often try to express our intuition about important /interesting element using an appropriate measure (index, weight)following the scheme
larger is the measure value of an element,more important / interesting is this element
Too often, in analysis of networks, researchers uncritically pick somemeasure from the literature. For formal approach see Roberts.
It seems that the most important distinction between different nodeindices is based on the view/decision whether the network isconsidered directed or undirected.
22 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
. . . Important nodes in network
This gives us two main types of indices:
• networks containing directed links (we replace edges by pairs ofopposite arcs): measures of importance; with two subgroups:measures of influence, based on out-going arcs; and measures ofsupport, based on incoming arcs;
• measures of centrality, based on all links.
For undirected networks all three types of measures coincide.If we change the direction of all arcs (replace the relation with itsinverse relation) the measure of influence becomes a measure ofsupport, and vice versa.
23 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
. . . Important nodes in network
The real meaning of measure of importance depends on the relationdescribed by a network. For example the most ’important’ person forthe relation ’ doesn’t like to work with ’ is in fact the leastpopular person.
Removal of an important node/link from a network produces asubstantial change in structure/functioning of the network.
24 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Normalization
Let p : V → R be an index in network N = (V,L, p). If we want tocompare indices p over different networks we have to make themcomparable. Usually we try to achieve this by normalization of p.Let N ∈ N(V), where N(V) is a selected family of networks over thesame set of nodes V ,
pmax = maxN∈N(V)
maxv∈V
pN (v) and pmin = minN∈N(V)
minv∈V
pN (v)
then we define the normalized index as
p′(v) =p(v)− pmin
pmax − pmin∈ [0, 1]
25 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Degrees
The simplest index are the degrees of nodes. Since for simplenetworks degmin = 0 and degmax = n − 1, the correspondingnormalized indices are
centrality deg′(v) =deg(v)
n − 1and similary
support indeg′(v) =indeg(v)
n
influence outdeg′(v) =outdeg(v)
nInstead of degrees in original network we can consider also thedegrees with respect to the reachability relation (transitive closure).
Network/Create Partition/Degree
Network/Create Vector/Centrality/Degree
Network/Create Vector/Centrality/Proximity Prestige
26 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Closeness
Most indices are based on the distance d(u, v) between vertices in anetwork N = (V,L). Two such indices areradius r(v) = maxu∈V d(v , u)total closeness S(v) =
∑u∈V d(v , u)
These two indices are measures of influence – to get measures ofsupport we have to replace in definitions d(u, v) with d(v , u).If the network is not strongly connected rmax and Smax are equal to∞. Sabidussi (1966) introduced a related measure 1/S(v); or in itsnormalized form
closeness cl(v) =n − 1∑
u∈V d(v , u)D = maxu,v∈V d(v , u) is called the diameter of network.
Network/Create Vector/Centrality/Closeness
Network/Create New Network/Subnetwork with Paths/Info
on Diameter
27 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Betweeness
Important are also the nodes that can control the information flow inthe network. If we assume that this flow uses only the shortest paths(geodesics) we get a measure of betweeness (Anthonisse 1971,Freeman 1977)
b(v) =1
(n − 1)(n − 2)
∑u,t∈V:gu,t>0
u 6=v,t 6=v,u 6=t
gu,t(v)
gu,t
where gu,t is the number of geodesics from u to t; and gu,t(v) is thenumber of those among them that pass through node v .
For computation of geodesic matrix see Brandes.
Network/Create Vector/Centrality/Betweenness
28 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Padgett’s Florentine families
Acciaiuoli
Albizzi
Barbadori
Bischeri
Castellani
Ginori
Guadagni
Lamberteschi
Medici
Pazzi
Peruzzi
Ridolfi
Salviati
Strozzi
Tornabuoni
close between
1. Acciaiuoli 0.368421 0.000000
2. Albizzi 0.482759 0.212454
3. Barbadori 0.437500 0.093407
4. Bischeri 0.400000 0.104396
5. Castellani 0.388889 0.054945
6. Ginori 0.333333 0.000000
7. Guadagni 0.466667 0.254579
8. Lamberteschi 0.325581 0.000000
9. Medici 0.560000 0.521978
10. Pazzi 0.285714 0.000000
11. Peruzzi 0.368421 0.021978
12. Ridolfi 0.500000 0.113553
13. Salviati 0.388889 0.142857
14. Strozzi 0.437500 0.102564
15. Tornabuoni 0.482759 0.091575
29 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Hubs and authorities
To each node v of a network N = (V,L) we assign two values:quality of its content (authority) xv and quality of its references(hub) yv .A good authority is selected by good hubs; and good hub points togood authorities (see Kleinberg).
xv =∑
u:(u,v)∈L
yu and yv =∑
u:(v ,u)∈L
xu
Let W be a matrix of network N and x and y authority and hubvectors. Then we can write these two relations as x = WTy andy = Wx.We start with y = [1, 1, . . . , 1] and then compute new vectors x andy. After each step we normalize both vectors. We repeat this untilthey stabilize.We can show that this procedure converges. The limit vector x∗ isthe principal eigen vector of matrix WTW; and y∗ of matrix WWT .
30 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
. . . Hubs and authorities
Similar procedures are used in search engines on the web to evaluatethe importance of web pages.PageRank, PageRank / Google, HITS / AltaVista, SALSA, theory.
Network/Create New Network/Subnetwork with Paths/Info on DiameterNetwork/Create Vector/Centrality/ClosenessNetwork/Create Vector/Centrality/BetweenessNetwork/Create Vector/Centrality/Hubs-AuthoritiesNetwork/Create Vector/Centrality/Clustering Coefficients
Examples: Krebs, Krempl. World Cup 1998 in Paris, 22 nationalteams. A player from first country is playing in the second country.
There exist other measures based on eigen-values and eigen-vectors
such as Katz, Bonachich and Brandes. See also Borgatti.
31 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
. . . Hubs and authorities: football
ARG
AUT
BEL
BGR
BRA
CHE
CHL
CMR
COL
DEU
DNK
ESP
FRA
GBR
GRE
HRV
IRN
ITA JAM
JPN
KOR
MAR
MEX
NGA
NLD
NORPRT
PRY
ROM
SCO
TUN
TUR
USA
YUG
ZAF
Exporters (hubs)
ARG
AUT
BEL
BGR
BRA
CHE
CHL
CMR
COL
DEU
DNK
ESP
FRA
GBR
GRE
HRV
IRN
ITA JAM
JPN
KOR
MAR
MEX
NGA
NLD
NORPRT
PRY
ROM
SCO
TUN
TUR
USA
YUG
ZAF
Importers (authorities)
32 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Clustering
Let G = (V, E) be simple undirected graph. Clustering in node v isusually measured as a quotient between the number of links insubgraph G1(v) = G(N1(v)) induced by the neighbors of node v andthe number of links in the complete graph on these nodes:
C (v) =
2|L(G1(v))|
deg(v)(deg(v)− 1)deg(v) > 1
0 otherwise
We can consider also the size of node neighborhood by the followingcorrection
C1(v) =deg(v)
∆C (v)
where ∆ is the maximum degree in graph G. This measure attains itslargest value in nodes that belong to an isolated clique of size ∆.
Network/Create Vector/Clustering Coefficients/CC2
paper33 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
User defined indices
Xingqin Qi et al. defined in their paper Terrorist Networks, NetworkEnergy and Node Removal a new measure of centrality based onLaplacian energy – Laplacian centrality
L(v) = deg(v)(deg(v) + 1) + 2∑
u∈N(v)
deg(u)
select the networkNetwork/Create Vector/Centrality/Degree/AllOperations/Network+Vector/Neighbours/Sum/All [False]Vector/Transform/Multiply by [2]select the degree vector as Firstselect the degree vector as SecondVectors/Multiply (First*Second)Vectors/Add (First+Second)select the 2*sum on neighbors as SecondVectors/Add (First+Second)dispose auxiliary vectorsFile/Vector/Change Label [Laplace All centrality]
macro
34 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
Network centralization measures
Extremal approach: Let p : V → R be an index in networkN = (V,L). We introduce the quantities
p? = maxv∈V
p(v)
D =∑v∈V
(p? − p(v))
D? = max{D(N ) : N is a network on V}
Then we can define centralization with respect to p
Cp(N ) =D(N )
D?
Usually the most centralized graph is the star Sn and the leastcentralized is the complete graph Kn.
35 / 36
Connectivity
V. Batagelj
Connectivity
Condensation
Bow-tie
Otherconnectivities
Importantnodes
Closeness
Betweeness
Hubs andauthorities
Clustering
. . . Network centralization measures
Variational approach: The other approach is based on the variance.First we compute the average node centrality
p =1
n
∑v∈V
p(v)
and then define
Vp(N ) =1
n
∑v∈V
(p(v)− p)2
36 / 36