WEBVTT

1
00:00:00.160 --> 00:00:04.360
<v Speaker 1>Welcome curious minds to the deep dive. Ever think about

2
00:00:04.360 --> 00:00:07.000
<v Speaker 1>that six degrees of separation idea?

3
00:00:07.040 --> 00:00:09.759
<v Speaker 2>Oh yeah, the classic party game concept, right, that.

4
00:00:09.720 --> 00:00:12.759
<v Speaker 1>You might only be like a handful of handshakes away

5
00:00:12.759 --> 00:00:14.560
<v Speaker 1>from anyone else on earth. It sounds kind of wild,

6
00:00:14.919 --> 00:00:17.399
<v Speaker 1>but it actually hints at something deeper about how connected

7
00:00:17.440 --> 00:00:17.960
<v Speaker 1>everything is.

8
00:00:18.440 --> 00:00:20.879
<v Speaker 2>It really does. It's a great hook into the whole

9
00:00:20.920 --> 00:00:21.399
<v Speaker 2>world of.

10
00:00:21.359 --> 00:00:24.800
<v Speaker 1>Networks exactly, and that's what we're doing today, deep dive

11
00:00:24.879 --> 00:00:28.879
<v Speaker 1>into networks. It's such a fundamental idea. We're using Mark

12
00:00:28.960 --> 00:00:33.759
<v Speaker 1>Newman's Networks second Edition as our guide. It's pretty much

13
00:00:33.799 --> 00:00:34.799
<v Speaker 1>the standard.

14
00:00:34.359 --> 00:00:37.920
<v Speaker 2>Text, definitely the authority in the field comprehensive stuff.

15
00:00:38.119 --> 00:00:41.479
<v Speaker 1>Our goal here is to unpack what networks are, the

16
00:00:41.520 --> 00:00:45.320
<v Speaker 1>different kinds you find and some of their practically surprising features.

17
00:00:45.640 --> 00:00:48.640
<v Speaker 1>Basically give you a solid understanding of this field, maybe

18
00:00:48.679 --> 00:00:51.159
<v Speaker 1>even change how you see the connections all around you.

19
00:00:51.359 --> 00:00:52.880
<v Speaker 2>Yeah, those invisible threads.

20
00:00:53.200 --> 00:00:56.640
<v Speaker 1>Okay, so let's start simple. What is a network? Fundamentally?

21
00:00:56.920 --> 00:00:59.560
<v Speaker 1>The book says it's basically a collection of points joined

22
00:00:59.560 --> 00:01:02.280
<v Speaker 1>together in pairs by lines. Seems almost too simple.

23
00:01:02.520 --> 00:01:05.920
<v Speaker 2>It does sound simple, but that's its power. Those points

24
00:01:05.959 --> 00:01:08.799
<v Speaker 2>we call them nodes. Technically, and the lines are edges

25
00:01:09.120 --> 00:01:11.760
<v Speaker 2>and ages. Got it, And the beauty is it's abstract.

26
00:01:12.280 --> 00:01:17.560
<v Speaker 2>Nodes can be people, computers, airports, brain cells, anything. Really,

27
00:01:17.879 --> 00:01:19.959
<v Speaker 2>Edges are just the connections between them.

28
00:01:20.120 --> 00:01:23.359
<v Speaker 1>So why is this simple idea so important across like

29
00:01:23.760 --> 00:01:24.879
<v Speaker 1>all kinds of science.

30
00:01:25.400 --> 00:01:28.640
<v Speaker 2>Because if you can model a system, any system, as

31
00:01:28.640 --> 00:01:32.920
<v Speaker 2>a network, suddenly you have this huge toolkit mathematical tools,

32
00:01:33.000 --> 00:01:37.879
<v Speaker 2>computational methods, hundreds of ways to analyze it, understand it,

33
00:01:37.920 --> 00:01:38.599
<v Speaker 2>predict its.

34
00:01:38.439 --> 00:01:41.000
<v Speaker 1>Behavior, like a universal language for complex stuff.

35
00:01:41.040 --> 00:01:43.439
<v Speaker 2>Pretty much. Yeah, it helps make sense of things that

36
00:01:43.519 --> 00:01:45.640
<v Speaker 2>might otherwise just seem like a tangled mess.

37
00:01:45.799 --> 00:01:47.840
<v Speaker 1>Okay, so let's talk types. Where do we see these

38
00:01:47.840 --> 00:01:51.079
<v Speaker 1>in the real world. The book mentions technological networks first.

39
00:01:50.959 --> 00:01:55.079
<v Speaker 2>Right, the physical infrastructure, think about the Internet, cables, routers,

40
00:01:55.799 --> 00:01:59.480
<v Speaker 2>or the power grid, telephone lines though maybe less relevant.

41
00:01:59.079 --> 00:02:03.599
<v Speaker 1>Now still, and transportation right, roads, railways, flight.

42
00:02:03.400 --> 00:02:08.479
<v Speaker 2>Paths, exactly, all physical systems connecting points. The Internet structure

43
00:02:08.520 --> 00:02:10.919
<v Speaker 2>is fascinating. Actually, it's got the sort of three layered shape,

44
00:02:10.960 --> 00:02:14.000
<v Speaker 2>three layers, like how we've got the backbone, these super

45
00:02:14.120 --> 00:02:18.680
<v Speaker 2>high capacity lines and routers crisscrossing countries, continents. Then you

46
00:02:18.680 --> 00:02:21.719
<v Speaker 2>have the big network backbone providers like telecoms or governments

47
00:02:21.800 --> 00:02:22.879
<v Speaker 2>managing large chunks.

48
00:02:23.000 --> 00:02:24.280
<v Speaker 1>Okay, but here's the kicker.

49
00:02:24.759 --> 00:02:29.240
<v Speaker 2>There's no single boss, no central control. The Internet Engineering

50
00:02:29.240 --> 00:02:34.639
<v Speaker 2>Task Force sets standards protocols, but anyone can essentially build

51
00:02:34.680 --> 00:02:35.759
<v Speaker 2>onto it. It just grew.

52
00:02:36.080 --> 00:02:38.599
<v Speaker 1>That sounds kind of chaotic. Does that make it fragile

53
00:02:39.000 --> 00:02:39.599
<v Speaker 1>or stronger?

54
00:02:39.879 --> 00:02:43.479
<v Speaker 2>Both? In a way. It's resilient to random failures. If

55
00:02:43.479 --> 00:02:47.280
<v Speaker 2>one router dies, data usually finds another path. We can

56
00:02:47.319 --> 00:02:49.039
<v Speaker 2>map this using tools like trice route.

57
00:02:49.120 --> 00:02:50.879
<v Speaker 1>I think I've used that pinging addresses.

58
00:02:51.039 --> 00:02:54.400
<v Speaker 2>Yeah, basically seeing the hops or by looking at routing

59
00:02:54.400 --> 00:02:57.879
<v Speaker 2>tables in what are called autonomous systems big networks managed

60
00:02:57.879 --> 00:03:01.000
<v Speaker 2>by ISPs, for example. But because it's decent centralized, coordinated

61
00:03:01.000 --> 00:03:04.240
<v Speaker 2>defense against attacks can be tricky and unlike some other networks,

62
00:03:04.280 --> 00:03:06.960
<v Speaker 2>Internet nodes, the actual computers and routers are mostly fixed

63
00:03:07.000 --> 00:03:08.080
<v Speaker 2>in physical locations.

64
00:03:08.400 --> 00:03:11.560
<v Speaker 1>Right, my laptop is here. Okay, beyond physical stuff, what

65
00:03:11.639 --> 00:03:14.400
<v Speaker 1>about information networks? That sounds more abstract.

66
00:03:14.680 --> 00:03:18.360
<v Speaker 2>Totally abstract, but hugely important. The Worldwide Web is the

67
00:03:18.360 --> 00:03:22.240
<v Speaker 2>prime example. Web Pages are nodes. Hyperlinks are the directed edges.

68
00:03:22.439 --> 00:03:23.240
<v Speaker 2>You click one.

69
00:03:23.120 --> 00:03:24.879
<v Speaker 1>Way doesn't always mean there's a link back.

70
00:03:24.960 --> 00:03:28.639
<v Speaker 2>Yeah, exactly. Or think about academic papers citation networks. One

71
00:03:28.639 --> 00:03:31.560
<v Speaker 2>paper citing another show how knowledge builds.

72
00:03:31.240 --> 00:03:35.360
<v Speaker 1>Over time like a family tree of ideas sort of.

73
00:03:35.599 --> 00:03:38.199
<v Speaker 2>Yeah, And you can even have cau citation networks where

74
00:03:38.240 --> 00:03:40.960
<v Speaker 2>two papers are linked. If another paper cites both of

75
00:03:41.039 --> 00:03:44.919
<v Speaker 2>them often means they're related. Even a simple keyword index

76
00:03:44.960 --> 00:03:47.879
<v Speaker 2>in a book is a type of information network linking

77
00:03:47.960 --> 00:03:48.840
<v Speaker 2>terms to pages.

78
00:03:49.120 --> 00:03:52.000
<v Speaker 1>Okay, interesting. Then there are social networks, and you mean

79
00:03:52.039 --> 00:03:55.000
<v Speaker 1>more than just like scrolling through Instagram.

80
00:03:54.520 --> 00:03:58.439
<v Speaker 2>Right, Oh, definitely. Scientifically, it's any network where people or

81
00:03:58.479 --> 00:04:00.879
<v Speaker 2>groups of people are the noes and the edges are

82
00:04:00.919 --> 00:04:06.000
<v Speaker 2>some kind of social tie, friendship, family, colleagues, who talks

83
00:04:06.039 --> 00:04:06.599
<v Speaker 2>to whom.

84
00:04:06.439 --> 00:04:09.479
<v Speaker 1>So actual human connection mapped out precisely.

85
00:04:09.639 --> 00:04:13.840
<v Speaker 2>Sociologists were doing this way before the Internet, using surveys records.

86
00:04:14.400 --> 00:04:17.920
<v Speaker 2>Now online platforms give us massive amounts of data, which is,

87
00:04:18.079 --> 00:04:22.240
<v Speaker 2>you know, both powerful and raises privacy questions, but it

88
00:04:22.319 --> 00:04:24.600
<v Speaker 2>lets us see the structure of society.

89
00:04:24.879 --> 00:04:28.160
<v Speaker 1>And the last big category is biological networks. That sounds

90
00:04:28.160 --> 00:04:29.040
<v Speaker 1>like it goes really deep.

91
00:04:29.160 --> 00:04:32.519
<v Speaker 2>It does, right down to the cellular level. Think about metabolic.

92
00:04:32.040 --> 00:04:34.079
<v Speaker 1>Networks, metabolic like digestion.

93
00:04:34.439 --> 00:04:38.079
<v Speaker 2>Fundamentally, yes, it's the network of chemical reactions happening in

94
00:04:38.120 --> 00:04:41.920
<v Speaker 2>your cells. The chemicals or metabolites are nodes and the

95
00:04:41.959 --> 00:04:45.480
<v Speaker 2>reactions transforming them are directed edges. It's like a ridiculously

96
00:04:45.560 --> 00:04:48.800
<v Speaker 2>detailed city map of your biochemistry. Wow. Then you have

97
00:04:48.959 --> 00:04:52.800
<v Speaker 2>networks of proteins interacting genes, regulating other genes, and of

98
00:04:52.839 --> 00:04:54.399
<v Speaker 2>course the brain neural.

99
00:04:54.240 --> 00:04:55.480
<v Speaker 1>Networks, ultimate network.

100
00:04:55.519 --> 00:04:59.639
<v Speaker 2>Maybe arguably, neurons are nodes with inputs, dendrites and outputs

101
00:04:59.639 --> 00:05:03.360
<v Speaker 2>exon the connections happen at synapses, these tiny gaps and

102
00:05:03.439 --> 00:05:06.560
<v Speaker 2>the strength of those connections can change. That's learning, Basically.

103
00:05:06.639 --> 00:05:09.040
<v Speaker 1>How does scientists even map that tricky stuff?

104
00:05:09.240 --> 00:05:12.759
<v Speaker 2>Sometimes using tracers like special dyes or viruses that hop

105
00:05:12.800 --> 00:05:16.839
<v Speaker 2>across synapses, or MRI scans for a bigger picture, though

106
00:05:16.839 --> 00:05:20.199
<v Speaker 2>that doesn't show individual neuron links. But yeah, the idea

107
00:05:20.240 --> 00:05:24.759
<v Speaker 2>is that consciousness thought, it emerges from this incredibly complex

108
00:05:24.879 --> 00:05:29.240
<v Speaker 2>web of connections. Life is networks.

109
00:05:28.920 --> 00:05:35.040
<v Speaker 1>Kind of mind blowing. Cells, websites, friends, brains, all networks.

110
00:05:35.319 --> 00:05:37.519
<v Speaker 1>And you mentioned they share common features like that six

111
00:05:37.600 --> 00:05:38.079
<v Speaker 1>degrees thing.

112
00:05:38.279 --> 00:05:41.240
<v Speaker 2>Exactly. That's the small world effect. It's wild in so

113
00:05:41.279 --> 00:05:44.639
<v Speaker 2>many different types of networks. The average pathlength, the number

114
00:05:44.639 --> 00:05:46.519
<v Speaker 2>of steps to get from any node to any other

115
00:05:46.600 --> 00:05:48.439
<v Speaker 2>node is surprisingly short.

116
00:05:48.360 --> 00:05:51.439
<v Speaker 1>Even in massive networks like millions of nodes.

117
00:05:51.279 --> 00:05:54.639
<v Speaker 2>Yep, often just six, seven, maybe a dozen steps. Stanley

118
00:05:54.680 --> 00:05:57.519
<v Speaker 2>Milgram's experiment in the sixties mailing letters, that's the one.

119
00:05:57.600 --> 00:06:00.120
<v Speaker 2>It showed this empirically. But what's maybe even more or

120
00:06:00.160 --> 00:06:03.120
<v Speaker 2>interesting isn't just that short paths exist. It's that people

121
00:06:03.120 --> 00:06:06.000
<v Speaker 2>are somehow really good at finding them, even without a map.

122
00:06:06.160 --> 00:06:09.040
<v Speaker 1>How does that even work? If I don't know the person?

123
00:06:09.279 --> 00:06:11.000
<v Speaker 1>How do I pick the right friend of a friend

124
00:06:11.000 --> 00:06:12.079
<v Speaker 1>to pass the message to?

125
00:06:12.519 --> 00:06:16.319
<v Speaker 2>That's the Puzzlemilgrim highlighted. It suggests the network isn't just random.

126
00:06:16.399 --> 00:06:19.240
<v Speaker 2>It has a specific structure, probably a miss of tight

127
00:06:19.279 --> 00:06:22.000
<v Speaker 2>local clusters and a few random long range links that

128
00:06:22.040 --> 00:06:26.720
<v Speaker 2>helps us navigate our intuition. Using local cues like geography

129
00:06:26.879 --> 00:06:29.839
<v Speaker 2>or occupation seems surprisingly effective.

130
00:06:29.959 --> 00:06:32.839
<v Speaker 1>So the network structure itself guides us. Okay, let's circle

131
00:06:32.839 --> 00:06:36.560
<v Speaker 1>back to some specifics we talked Internet structure. How does

132
00:06:36.560 --> 00:06:40.120
<v Speaker 1>the old telephone network compare. It's been around much longer.

133
00:06:40.120 --> 00:06:44.040
<v Speaker 2>Right over a century. Structurally, as overall shape, its topology

134
00:06:44.120 --> 00:06:46.439
<v Speaker 2>didn't change radically For a long time. It was very

135
00:06:46.759 --> 00:06:50.000
<v Speaker 2>geographically based. People mostly called locally makes sense, but the

136
00:06:50.040 --> 00:06:54.839
<v Speaker 2>technology underneath completely transformed the mainlines. The trunk lines are

137
00:06:54.879 --> 00:06:58.120
<v Speaker 2>now digital packet switched. Often they run on the exact

138
00:06:58.120 --> 00:07:01.360
<v Speaker 2>same fiber optic cables as the Internet. Your voice call

139
00:07:01.519 --> 00:07:04.279
<v Speaker 2>might actually be hopping over Internet infrastructure for most of

140
00:07:04.319 --> 00:07:05.680
<v Speaker 2>its journey, So only.

141
00:07:05.480 --> 00:07:07.360
<v Speaker 1>That last bit of copper wire to the house is

142
00:07:07.399 --> 00:07:08.839
<v Speaker 1>maybe old school pretty much.

143
00:07:08.879 --> 00:07:11.720
<v Speaker 2>Yeah, and even that's changing fast with fiber to the home.

144
00:07:12.120 --> 00:07:14.639
<v Speaker 2>The function is similar, but the tech is converged.

145
00:07:15.000 --> 00:07:19.519
<v Speaker 1>What about transportation, You mentioned roads, air travel, Any non

146
00:07:19.560 --> 00:07:21.560
<v Speaker 1>obvious network insights there, well.

147
00:07:21.439 --> 00:07:24.839
<v Speaker 2>Yeah, consider railways. You could just map stations as nodes

148
00:07:24.879 --> 00:07:28.120
<v Speaker 2>and tracks as edges. Simple enough, okay, But studies like

149
00:07:28.160 --> 00:07:30.480
<v Speaker 2>one on the Indian rail network pointed out something crucial.

150
00:07:31.319 --> 00:07:33.920
<v Speaker 2>Passengers don't just care about track connections, They care about

151
00:07:33.920 --> 00:07:36.639
<v Speaker 2>staying on the same train. Changing trains is a big deal.

152
00:07:36.839 --> 00:07:39.319
<v Speaker 1>Ah, right, the hassle factor exactly.

153
00:07:39.480 --> 00:07:42.360
<v Speaker 2>So a more useful network model might be what's called

154
00:07:42.360 --> 00:07:43.560
<v Speaker 2>a bipartite network.

155
00:07:44.000 --> 00:07:45.199
<v Speaker 1>Bipartite two parts.

156
00:07:45.240 --> 00:07:48.240
<v Speaker 2>Yeah, you have two types of nodes, say, stations and

157
00:07:48.279 --> 00:07:51.399
<v Speaker 2>specific train routes, and edge only connects the station to

158
00:07:51.480 --> 00:07:54.000
<v Speaker 2>a route. If that train stops there, it captures the

159
00:07:54.000 --> 00:07:55.959
<v Speaker 2>single journey possibility much better.

160
00:07:56.279 --> 00:07:59.759
<v Speaker 1>That's clever. It models the user experience better precisely. Yeah.

161
00:08:00.000 --> 00:08:02.720
<v Speaker 2>Wh In back in biology we mentioned metabolic and neural networks.

162
00:08:02.920 --> 00:08:07.519
<v Speaker 2>It's all about mapping those connections chemicals to reactions, neurons

163
00:08:07.560 --> 00:08:11.000
<v Speaker 2>to synapses to understand how the system functions, how life

164
00:08:11.079 --> 00:08:13.079
<v Speaker 2>actually works, or how thought emerges.

165
00:08:13.399 --> 00:08:16.360
<v Speaker 1>Okay, so we have all these different networks to really

166
00:08:16.399 --> 00:08:20.879
<v Speaker 1>study the mathematically. What core concepts do we need? What

167
00:08:20.920 --> 00:08:22.160
<v Speaker 1>are the essential tools?

168
00:08:22.279 --> 00:08:26.439
<v Speaker 2>Good question. We need some basic vocabulary first. Often we

169
00:08:26.519 --> 00:08:30.000
<v Speaker 2>deal with simple networks, no loops, where a node connects

170
00:08:30.040 --> 00:08:33.320
<v Speaker 2>to itself and only one edge between any two nodes.

171
00:08:33.120 --> 00:08:34.399
<v Speaker 1>Makes sense, keep it clean.

172
00:08:34.759 --> 00:08:38.600
<v Speaker 2>But sometimes you need multigraphs which do allow multiple edges,

173
00:08:38.600 --> 00:08:41.039
<v Speaker 2>like maybe several different types of relationships between two people,

174
00:08:41.159 --> 00:08:45.039
<v Speaker 2>or multiple cables between two routers. Okay, then direction matters.

175
00:08:45.120 --> 00:08:48.840
<v Speaker 2>Directive networks have edges with arrows like the weblinks or

176
00:08:49.000 --> 00:08:51.360
<v Speaker 2>food webs, energy flows from the grass to the rabbit,

177
00:08:51.480 --> 00:08:54.120
<v Speaker 2>not the other way, or citations.

178
00:08:53.519 --> 00:08:55.679
<v Speaker 1>Paper A sites, paper B one direction.

179
00:08:56.080 --> 00:08:58.559
<v Speaker 2>Got it, and we touched on bipart type networks. Those

180
00:08:58.559 --> 00:09:02.000
<v Speaker 2>two distinct node types. Actors and movies. They're in edges

181
00:09:02.080 --> 00:09:04.000
<v Speaker 2>only go between types, not within.

182
00:09:04.120 --> 00:09:07.320
<v Speaker 1>Actors connect to movies. Movies connect to actors. No actor

183
00:09:07.320 --> 00:09:09.279
<v Speaker 1>to actor edge in that basic model.

184
00:09:09.240 --> 00:09:12.440
<v Speaker 2>Right, But from that you can create a one mode

185
00:09:12.519 --> 00:09:15.000
<v Speaker 2>projection like connect two actors if they appeared in the

186
00:09:15.039 --> 00:09:18.600
<v Speaker 2>same movie. Suddenly you have an actor actor network derived

187
00:09:18.720 --> 00:09:22.240
<v Speaker 2>from the bipartite one. It reveals indirect connections.

188
00:09:22.360 --> 00:09:24.919
<v Speaker 1>Okay, that makes sense. So once we have the network structure,

189
00:09:24.960 --> 00:09:27.759
<v Speaker 1>how do we figure out which nodes are important?

190
00:09:28.240 --> 00:09:32.720
<v Speaker 2>Central centrality key concept, and there are different ways to

191
00:09:32.759 --> 00:09:35.480
<v Speaker 2>measure importance depending on what you mean. The simplest is

192
00:09:35.639 --> 00:09:37.120
<v Speaker 2>degree centrality.

193
00:09:37.240 --> 00:09:38.200
<v Speaker 1>Just counting connections.

194
00:09:38.320 --> 00:09:42.080
<v Speaker 2>Yep, how many edges does a node have indirected networks?

195
00:09:42.080 --> 00:09:46.840
<v Speaker 2>You'd count incoming edges in degree and outgoing edges out

196
00:09:47.039 --> 00:09:50.320
<v Speaker 2>degree separately. Easy to calculate, but maybe not always the

197
00:09:50.320 --> 00:09:51.600
<v Speaker 2>whole story exactly.

198
00:09:51.960 --> 00:09:54.039
<v Speaker 1>Sometimes it's not just how many people you know, but

199
00:09:54.120 --> 00:09:57.200
<v Speaker 1>who you know that leads to eigenvector centrality.

200
00:09:57.279 --> 00:09:59.440
<v Speaker 2>Eigenvector sounds complex.

201
00:09:59.559 --> 00:10:03.039
<v Speaker 1>The idea simple, though, your importance increases if you're connected

202
00:10:03.080 --> 00:10:06.759
<v Speaker 1>to other important notes. It's recursive, like that's saying having

203
00:10:06.759 --> 00:10:09.039
<v Speaker 1>one friends the president makes you more important than having

204
00:10:09.120 --> 00:10:09.720
<v Speaker 1>ten friends.

205
00:10:10.200 --> 00:10:14.200
<v Speaker 2>Nobody knows right influence by association. Then there's cats centrality.

206
00:10:14.399 --> 00:10:17.120
<v Speaker 2>It's similar, but it gives every node a little bit

207
00:10:17.159 --> 00:10:21.120
<v Speaker 2>of baseline importance automatically, and that importance can flow outwards,

208
00:10:21.519 --> 00:10:23.639
<v Speaker 2>but it diminishes with distance, so.

209
00:10:23.600 --> 00:10:25.480
<v Speaker 1>Even isolated nodes have some score.

210
00:10:25.639 --> 00:10:28.879
<v Speaker 2>Yeah, and it accounts for influence spreading through longer paths.

211
00:10:29.279 --> 00:10:32.159
<v Speaker 2>Page erank the Google algorithm is a variation.

212
00:10:31.840 --> 00:10:33.919
<v Speaker 1>Of this new Google would come up right.

213
00:10:34.240 --> 00:10:37.000
<v Speaker 2>It ranks pages based on links from other important pages,

214
00:10:37.039 --> 00:10:39.679
<v Speaker 2>but it cleverly adjusts for the number of links the

215
00:10:39.720 --> 00:10:42.759
<v Speaker 2>source page sends out, so a link from a focused

216
00:10:42.759 --> 00:10:45.440
<v Speaker 2>important page counts more than one link from a huge

217
00:10:45.519 --> 00:10:49.480
<v Speaker 2>hub like say Wikipedia's front page, which links everywhere.

218
00:10:49.639 --> 00:10:51.320
<v Speaker 1>Prevents those hubs from having too much.

219
00:10:51.159 --> 00:10:55.600
<v Speaker 2>Influence exactly, and one more key one between highness centrality.

220
00:10:55.720 --> 00:10:58.759
<v Speaker 2>This measures something different, okay. It measures how often a

221
00:10:58.840 --> 00:11:01.799
<v Speaker 2>node lies on the shortest path between other pairs of nodes.

222
00:11:01.960 --> 00:11:03.960
<v Speaker 1>So it's about being a bridge precisely.

223
00:11:04.000 --> 00:11:06.799
<v Speaker 2>You could have a node with only two connections low degree,

224
00:11:07.159 --> 00:11:09.759
<v Speaker 2>but if it's the only link between two large communities

225
00:11:09.759 --> 00:11:12.120
<v Speaker 2>and the network, it has a very high between us.

226
00:11:12.559 --> 00:11:16.039
<v Speaker 2>It's critical for information flow or connection between those groups.

227
00:11:16.320 --> 00:11:19.279
<v Speaker 1>Like that one person who knows people from two totally

228
00:11:19.279 --> 00:11:20.279
<v Speaker 1>different clicks.

229
00:11:19.960 --> 00:11:22.559
<v Speaker 2>Perfect analogy. They control the flow between them.

230
00:11:22.879 --> 00:11:27.080
<v Speaker 1>Fascinating. Okay, so different ways to be central. What about resilience?

231
00:11:27.399 --> 00:11:31.799
<v Speaker 1>What happens when networks break, parts fail, nodes get removed.

232
00:11:32.039 --> 00:11:35.240
<v Speaker 2>That's where percolation theory comes in. It studies how robust

233
00:11:35.279 --> 00:11:39.039
<v Speaker 2>the network is. And there's a really interesting finding related

234
00:11:39.080 --> 00:11:42.840
<v Speaker 2>to specific network types called scale free networks.

235
00:11:42.960 --> 00:11:45.279
<v Speaker 1>Scale free what defines those.

236
00:11:45.480 --> 00:11:49.039
<v Speaker 2>They have what's called a power law degree distribution. Basically,

237
00:11:49.159 --> 00:11:51.799
<v Speaker 2>most nodes have very few connections, but there are a

238
00:11:51.840 --> 00:11:55.159
<v Speaker 2>few nodes hubs that have a massive number of connections.

239
00:11:55.200 --> 00:11:57.840
<v Speaker 2>The Internet is often cited as an example. Maybe social

240
00:11:57.879 --> 00:11:59.120
<v Speaker 2>networks too, lots.

241
00:11:58.960 --> 00:12:02.159
<v Speaker 1>Of regular users, few influencers or major sites.

242
00:12:02.320 --> 00:12:05.200
<v Speaker 2>Kind of Yeah. Now, the surprising thing is these networks

243
00:12:05.200 --> 00:12:08.639
<v Speaker 2>are actually very resilient to random failures. You can knock

244
00:12:08.679 --> 00:12:11.360
<v Speaker 2>out a bunch of random nodes, and the main connected

245
00:12:11.360 --> 00:12:14.639
<v Speaker 2>part of the network, the giant component, usually stays connected.

246
00:12:14.720 --> 00:12:17.080
<v Speaker 2>It can take a hit, it can and it's a bit,

247
00:12:17.399 --> 00:12:21.679
<v Speaker 2>but they are extremely vulnerable to targeted attacks. Meaning if

248
00:12:21.679 --> 00:12:24.919
<v Speaker 2>you specifically go after those few high degree hubs and

249
00:12:25.000 --> 00:12:29.480
<v Speaker 2>take them out, the whole network can shatter Removing just

250
00:12:29.559 --> 00:12:33.039
<v Speaker 2>a tiny fraction of the most connected nose can completely

251
00:12:33.080 --> 00:12:37.679
<v Speaker 2>disconnect the giant component, breaking the network into isolated fragments.

252
00:12:37.759 --> 00:12:43.360
<v Speaker 1>So strong against accidents, weak against deliberate attacks on key points.

253
00:12:43.440 --> 00:12:47.799
<v Speaker 2>That's the takeaway for scale free networks. Structure dictates resilience

254
00:12:48.159 --> 00:12:49.159
<v Speaker 2>and vulnerability.

255
00:12:49.399 --> 00:12:52.600
<v Speaker 1>Wow. Okay, what a journey we've taken. We went from

256
00:12:52.639 --> 00:12:55.080
<v Speaker 1>the simple points and lines definition, yeah, all the way

257
00:12:55.080 --> 00:12:58.080
<v Speaker 1>through tech info social biological network Yeah.

258
00:12:58.120 --> 00:13:00.200
<v Speaker 2>Covered a lot of ground how we map them and

259
00:13:00.200 --> 00:13:01.919
<v Speaker 2>measure things like centrality, right.

260
00:13:01.919 --> 00:13:06.080
<v Speaker 1>Degree eigenvector between creedness, and then thinking about how they

261
00:13:06.080 --> 00:13:08.840
<v Speaker 1>hold up under stress. That whole scale free resilience and

262
00:13:08.919 --> 00:13:09.759
<v Speaker 1>vulnerability thing.

263
00:13:09.879 --> 00:13:11.559
<v Speaker 2>It really shows how these concepts.

264
00:13:11.240 --> 00:13:14.559
<v Speaker 1>Apply everywhere, absolutely, And what this deep dive really highlights

265
00:13:14.559 --> 00:13:17.960
<v Speaker 1>for you listening is how interconnected everything is. It's not

266
00:13:18.039 --> 00:13:20.840
<v Speaker 1>just a metaphor the algorithm shaping your web searches, the

267
00:13:20.879 --> 00:13:24.679
<v Speaker 1>way diseases might spread, how your own brain works. It's

268
00:13:24.720 --> 00:13:25.399
<v Speaker 1>all networks.

269
00:13:25.720 --> 00:13:29.320
<v Speaker 2>Understanding even the basics helps you see those hidden structures

270
00:13:29.639 --> 00:13:32.679
<v Speaker 2>think more critically about complex systems all around us.

271
00:13:32.919 --> 00:13:35.559
<v Speaker 1>Definitely, So here's something to chew on. A final thought

272
00:13:35.559 --> 00:13:39.000
<v Speaker 1>from network science. It's called the friendship paradox.

273
00:13:39.120 --> 00:13:40.200
<v Speaker 2>Ah, that's a good one.

274
00:13:40.279 --> 00:13:43.799
<v Speaker 1>Statistically, on average, your friends have more friends than you do.

275
00:13:44.399 --> 00:13:47.799
<v Speaker 2>Sounds weird, but it's mathematically true for most social networks.

276
00:13:48.320 --> 00:13:51.759
<v Speaker 1>Think about why that might be. Why would sampling nodes

277
00:13:51.879 --> 00:13:55.279
<v Speaker 1>via edges your friends tend to land you on higher

278
00:13:55.320 --> 00:13:58.480
<v Speaker 1>degree nodes. What does that say about our own perception

279
00:13:58.679 --> 00:14:01.159
<v Speaker 1>versus the reality of the net work we're in. Maybe

280
00:14:01.240 --> 00:14:03.200
<v Speaker 1>our local view isn't the whole picture.