Who to watch when? Strategic observation in the inverse ising problem
Who to watch when? Strategic observation in the inverse ising problem
In this paper, we investigate the problem of inferring the network coupling strengths from partially observed time series data in an Ising model on scale-free networks. By assuming that only a certain fraction of observations for spin states are available, we study how an observer, who wants to maximise the accuracy of the network inference, should distribute a limited number of observations. Along with the benchmark case of randomly-chosen hidden nodes, we propose degree-dependent heuristics for observation allocations.
We observe two regimes for the best observation strategies based on varying amounts of missing data. If only a small proportion of data cannot be observed, then one should focus on the observation of the states of low-degree nodes. Otherwise, if a large number of states cannot be observed, allocating more observations to the high-degree nodes is preferable.
Network inference, Inverse Ising model, Complex networks
1-4
Cai, Zhongqi
b3ce4c1b-e545-4a86-9592-960542756e14
Gerding, Enrico
d9e92ee5-1a8c-4467-a689-8363e7743362
Brede, Markus
bbd03865-8e0b-4372-b9d7-cd549631f3f7
Cai, Zhongqi
b3ce4c1b-e545-4a86-9592-960542756e14
Gerding, Enrico
d9e92ee5-1a8c-4467-a689-8363e7743362
Brede, Markus
bbd03865-8e0b-4372-b9d7-cd549631f3f7
Cai, Zhongqi, Gerding, Enrico and Brede, Markus
(2023)
Who to watch when? Strategic observation in the inverse ising problem.
French Regional Conference on Complex Systems<br/>FRCCS 2023, Le Havre, France, Le Havre, France.
31 May - 02 Jun 2023.
.
Record type:
Conference or Workshop Item
(Paper)
Abstract
In this paper, we investigate the problem of inferring the network coupling strengths from partially observed time series data in an Ising model on scale-free networks. By assuming that only a certain fraction of observations for spin states are available, we study how an observer, who wants to maximise the accuracy of the network inference, should distribute a limited number of observations. Along with the benchmark case of randomly-chosen hidden nodes, we propose degree-dependent heuristics for observation allocations.
We observe two regimes for the best observation strategies based on varying amounts of missing data. If only a small proportion of data cannot be observed, then one should focus on the observation of the states of low-degree nodes. Otherwise, if a large number of states cannot be observed, allocating more observations to the high-degree nodes is preferable.
Text
FRCCS_2023_paper_7001 (2)
- Version of Record
Restricted to Repository staff only
Request a copy
More information
Accepted/In Press date: 2023
e-pub ahead of print date: 31 May 2023
Venue - Dates:
French Regional Conference on Complex Systems<br/>FRCCS 2023, Le Havre, France, Le Havre, France, 2023-05-31 - 2023-06-02
Keywords:
Network inference, Inverse Ising model, Complex networks
Identifiers
Local EPrints ID: 476321
URI: http://eprints.soton.ac.uk/id/eprint/476321
PURE UUID: aec0d1fd-590b-4b02-b0c6-51969ec2a77b
Catalogue record
Date deposited: 22 Jan 2025 17:43
Last modified: 23 Jan 2025 02:40
Export record
Contributors
Author:
Zhongqi Cai
Author:
Enrico Gerding
Author:
Markus Brede
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics