Automatic Discovery of Families of Network Generative Processes

  title={Automatic Discovery of Families of Network Generative Processes},
  author={Telmo Menezes and Camille Roth},
Designing plausible network models typically requires scholars to form a priori intuitions on the key drivers of network formation. Oftentimes, these intuitions are supported by the statistical estimation of a selection of network evolution processes which will form the basis of the model to be developed. Machine learning techniques have lately been introduced to assist the automatic discovery of generative models. These approaches may more broadly be described as "symbolic regression", where… 

Examining the variability in network populations and its role in generative models

The need for rethinking the way the way the authors evaluate the goodness-of-fit of new and existing network models and devising models that are capable of matching the variability of network populations when available is highlighted.


The focus of this work is to introduce a reusable, generalized approach for the generation of synthetic social networks, using Symbolic regression to automatically train generation rules based on a network sample, instead of having to define rules a priori.

Symbolic regression of generative network models

This work has developed an approach to automatically detect realistic decentralised network growth models from empirical data, employing a machine learning technique inspired by natural selection and defining a unified formalism to describe such models as computer programs.

Genetic Programming for the Automatic Inference of Graph Models for Complex Networks

This paper proposes solutions to problems of an automated approach based on genetic programming and shows experimentally that the new system represents a significant improvement and is very capable of reproducing existing common graph models from even a single small initial network.

Automatic generation of graph models for complex networks by genetic programming

This paper represents the first exploration into the use of genetic programming for automating the discovery and algorithm design of graph models, representing a totally new approach with great interdisciplinary application potential.

Revealing the Hidden Language of Complex Networks

This work discovers that the interaction between a small number of roles, played by nodes in a network, can characterize a network's structure and also provide a clear real-world interpretation, and develops a framework for analysing and comparing networks, which outperforms all existing ones.

Generative models of the human connectome

Automatic Discovery of Agent Based Models: an Application to Social Anthropology

A methodology that applies a machine learning technique — genetic programming — to the problem of finding plausible generative models for complex networks and finds that this approach is capable of generating high quality results both in terms of network similarity and human readability of the programs.

Evolutionary modeling of a blog network

  • Telmo Menezes
  • Computer Science
    2011 IEEE Congress of Evolutionary Computation (CEC)
  • 2011
This paper proposes an approach based on an hybrid strategy, combining a generic behavioral template created by a human designer with a set of programs evolved using genetic programming, which shows that these clusters map to different behaviors, giving credence to the need for exogenous explanations for the dynamics of blog networks.

Action-based Modeling of Complex Networks

A robust new approach to Action-based Modeling is presented that creates a compact probabilistic model of a given target network, which can then be used to synthesize networks of arbitrary size and allows the user to consider an arbitrarily large set of structural characteristics during the generator design process.

Investigating Fitness Measures for the Automatic Construction of Graph Models

A proposed meta-analysis demonstrated how these network measures interacted when used together as classifiers to determine network, and thus model, (dis)similarity in a number of well-known network properties.

Taxonomies of networks from community structure.

A framework for constructing taxonomies of networks based on their structural similarities is introduced and applied to construct a taxonomy for 746 networks and it is demonstrated that the approach usefully identifies similar networks.