Back
2024INF016
Doctoral thesis

The dynamics of innovation : non-linear modelling of patent citations

  • 2024

PhD: Università della Svizzera italiana

Relational event modelling
Dynamic network modelling
Patent citation analysis
Deep learning
Non-linear modeling techniques
English This dissertation explores the development and application of advanced modeling techniques to analyze network dynamics, with a specific focus on patent citation networks. This study primarily centers on Relational Event Models (REMs), which have proven effective in modeling sequential interactions within networks but face significant limitations when applied to large datasets and complex nonlinear relationships. To address these challenges, this work develops extensions of the REM, including the Stochastic Gradient Relational Event Additive Model (STREAM) and the Deep Relational Event Additive Model (DREAM), both of which incorporate non-linear modeling techniques and deep learning approaches. The research presented here offers several key contributions. First, a method for computing textual similarity scores using embeddings from patent abstracts is proposed, demonstrating efficiency and effectiveness in capturing complex relationships within the citation profiles of patent data. Second, the dissertation provides a comprehensive review of REMs, highlighting their evolution and identifying areas for further development. The introduction of STREAM addresses the computational challenges of applying REMs to large-scale networks. In an application to patent citations, it reveals non-linear patterns in patent citation rates, particularly during periods of heightened technological innovation. Finally, DREAM leverages neural networks to model non-linear effects in large dynamic networks, offering a scalable and robust solution for analyzing large relational datasets with complex drivers. To demonstrate the high flexibility of these models, an application to the European interbank market is presented.
Collections
Language
  • English
Classification
Computer science and technology
License
License undefined
Open access status
green
Identifiers
Persistent URL
https://n2t.net/ark:/12658/srd1330484
Statistics
Document views: 26 File downloads:
  • 2024INF016.pdf: 76