Community detection in networks with unobserved edges Leto Peel - - PowerPoint PPT Presentation

community detection in networks with unobserved edges
SMART_READER_LITE
LIVE PREVIEW

Community detection in networks with unobserved edges Leto Peel - - PowerPoint PPT Presentation

Community detection in networks with unobserved edges Leto Peel Universit catholique de Louvain @PiratePeel Community detection Aim: partition the network according similarity of link structure Community detection Aim: partition the


slide-1
SLIDE 1

Community detection in networks with unobserved edges

Leto Peel Université catholique de Louvain @PiratePeel

slide-2
SLIDE 2

Community detection

Aim: partition the network according similarity of link structure

slide-3
SLIDE 3

Community detection

But we observe signals on nodes and no links! Aim: partition the network according similarity of link structure

slide-4
SLIDE 4

Motivating examples...

Identify assets whose prices vary coherently to better manage risk

slide-5
SLIDE 5

Motivating examples...

Identify assets whose prices vary coherently to better manage risk Identify regions of the brain to predict the onset of psychosis and learn about the ageing of the brain

slide-6
SLIDE 6

Motivating examples...

Identify assets whose prices vary coherently to better manage risk Identify regions of the brain to predict the onset of psychosis and learn about the ageing of the brain Identify climate zones to better understand factors afgecting our climate

slide-7
SLIDE 7

Is there really a network?

slide-8
SLIDE 8

Is there really a network?

We don’t have to directly observe something to believe it is true

slide-9
SLIDE 9

Common practise

  • Calculate pairwise correlations between signals (e.g. Pearson’s).
  • Threshold (and Binarize) the matrix of correlations.
  • Perform community detection on this (notional) network
slide-10
SLIDE 10

Problems

  • This procedure commonly invokes point-estimates at each step

– Does not capture the uncertainty of individual links

slide-11
SLIDE 11

Problems

  • This procedure commonly invokes point-estimates at each step

– Does not capture the uncertainty of individual links

  • Unclear how to include missing data.
  • No intrinsic/clear notion of the right number of communities.
slide-12
SLIDE 12

The signals we observe from many nodes are driven by a few latent factors

slide-13
SLIDE 13

Notion of a community is: a group of nodes that infmuenced similarly by the latent factors

The signals we observe from many nodes are driven by a few latent factors

slide-14
SLIDE 14

Observed time series Latent factor time series Factor loadings

slide-15
SLIDE 15

Community mean Community precision

slide-16
SLIDE 16

Generated Inferred Lower bound on the marginal likelihood (ELBO) Difgerence between Kgenerated and Kinferred

slide-17
SLIDE 17
slide-18
SLIDE 18
slide-19
SLIDE 19

US cities climate data

Koppen climate zones inferred climate zones

slide-20
SLIDE 20

What happened to the network?

  • Since we skip explicit interpretation of A our inference framework is

basically a Bayesian (time-series) clustering.

  • One can re-interpret AAT as a network, or interpret distances

between time-series in the latent-space as links in a network, but this is optional.

slide-21
SLIDE 21

EDGES? WHERE WE’RE GOING, WE DON’T NEED “EDGES”

slide-22
SLIDE 22

In collaboration with...

Renaud Lambiotte

Contact: leto.peel@uclouvain.be @PiratePeel

Till Hofgmann Nick Jones