-
Relating graph auto-encoders to linear models
Authors:
Solveig Klepper,
Ulrike von Luxburg
Abstract:
Graph auto-encoders are widely used to construct graph representations in Euclidean vector spaces. However, it has already been pointed out empirically that linear models on many tasks can outperform graph auto-encoders. In our work, we prove that the solution space induced by graph auto-encoders is a subset of the solution space of a linear map. This demonstrates that linear embedding models have…
▽ More
Graph auto-encoders are widely used to construct graph representations in Euclidean vector spaces. However, it has already been pointed out empirically that linear models on many tasks can outperform graph auto-encoders. In our work, we prove that the solution space induced by graph auto-encoders is a subset of the solution space of a linear map. This demonstrates that linear embedding models have at least the representational power of graph auto-encoders based on graph convolutional networks. So why are we still using nonlinear graph auto-encoders? One reason could be that actively restricting the linear solution space might introduce an inductive bias that helps improve learning and generalization. While many researchers believe that the nonlinearity of the encoder is the critical ingredient towards this end, we instead identify the node features of the graph as a more powerful inductive bias. We give theoretical insights by introducing a corresponding bias in a linear model and analyzing the change in the solution space. Our experiments are aligned with other empirical work on this question and show that the linear encoder can outperform the nonlinear encoder when using feature information.
△ Less
Submitted 30 November, 2023; v1 submitted 3 November, 2022;
originally announced November 2022.
-
Using Process Models to understand Security Standards
Authors:
Fabiola Moyón,
Daniel Méndez,
Kristian Beckers,
Sebastian Klepper
Abstract:
Many industrial software development processes today have to comply with security standards such as the IEC~62443-4-1. These standards, written in natural language, are ambiguous and complex to understand. This is especially true for non-security experts. Security practitioners thus invest much effort into comprehending standards and, later, into introducing them to development teams. However, our…
▽ More
Many industrial software development processes today have to comply with security standards such as the IEC~62443-4-1. These standards, written in natural language, are ambiguous and complex to understand. This is especially true for non-security experts. Security practitioners thus invest much effort into comprehending standards and, later, into introducing them to development teams. However, our experience in the industry shows that development practitioners might very well also read such standards, but nevertheless end up inviting experts for interpretation (or confirmation). Such a scenario is not in tune with current trends and needs of increasing velocity in continuous software engineering.
In this paper, we propose a tool-supported approach to make security standards more precise and easier to understand for both non-security as well as security experts by applying process models. This approach emerges from a large industrial company and encompasses so far the IEC62443-4-1 standard. We further present a case study with 16 industry practitioners showing how the approach improves communication between development and security compliance practitioners.
△ Less
Submitted 27 May, 2021;
originally announced May 2021.
-
How to Integrate Security Compliance Requirements with Agile Software Engineering at Scale?
Authors:
Fabiola Moyón,
Daniel Méndez Fernández,
Kristian Beckers,
Sebastian Klepper
Abstract:
Integrating security into agile software development is an open issue for research and practice. Especially in strongly regulated industries, complexity increases not only when scaling agile practices but also when aiming for compliance with security standards. To achieve security compliance in a large-scale agile context, we developed S2C-SAFe: An extension of the Scaled Agile Framework that is c…
▽ More
Integrating security into agile software development is an open issue for research and practice. Especially in strongly regulated industries, complexity increases not only when scaling agile practices but also when aiming for compliance with security standards. To achieve security compliance in a large-scale agile context, we developed S2C-SAFe: An extension of the Scaled Agile Framework that is compliant to the security standard IEC~62443-4-1 for secure product development.
In this paper, we present the framework and its evaluation by agile and security experts within Siemens' large-scale project ecosystem. We discuss benefits and limitations as well as challenges from a practitioners' perspective. Our results indicate that \ssafe contributes to successfully integrating security compliance with lean and agile development in regulated environments. We also hope to raise awareness for the importance and challenges of integrating security in the scope of Continuous Software Engineering.
△ Less
Submitted 27 May, 2021;
originally announced May 2021.
-
Clustering with Tangles: Algorithmic Framework and Theoretical Guarantees
Authors:
Solveig Klepper,
Christian Elbracht,
Diego Fioravanti,
Jakob Kneip,
Luca Rendsburg,
Maximilian Teegen,
Ulrike von Luxburg
Abstract:
Originally, tangles were invented as an abstract tool in mathematical graph theory to prove the famous graph minor theorem. In this paper, we showcase the practical potential of tangles in machine learning applications. Given a collection of cuts of any dataset, tangles aggregate these cuts to point in the direction of a dense structure. As a result, a cluster is softly characterized by a set of c…
▽ More
Originally, tangles were invented as an abstract tool in mathematical graph theory to prove the famous graph minor theorem. In this paper, we showcase the practical potential of tangles in machine learning applications. Given a collection of cuts of any dataset, tangles aggregate these cuts to point in the direction of a dense structure. As a result, a cluster is softly characterized by a set of consistent pointers. This highly flexible approach can solve clustering problems in various setups, ranging from questionnaires over community detection in graphs to clustering points in metric spaces. The output of our proposed framework is hierarchical and induces the notion of a soft dendrogram, which can help explore the cluster structure of a dataset. The computational complexity of aggregating the cuts is linear in the number of data points. Thus the bottleneck of the tangle approach is to generate the cuts, for which simple and fast algorithms form a sufficient basis. In our paper we construct the algorithmic framework for clustering with tangles, prove theoretical guarantees in various settings, and provide extensive simulations and use cases. Python code is available on github.
△ Less
Submitted 6 November, 2022; v1 submitted 25 June, 2020;
originally announced June 2020.
