Aindo Synthetic Data Platform 3.1 - Intuitive, innovative, in line with industry needs
We are proud to announce the release of the new version of the Aindo's Synthetic Data Platform, Aindo Synth. We worked closely with end-users in healthcare, finance, and insurance to determine key synthetic data needs. With the new release, we meet these needs with an innovative, easy to use and efficient platform.
Particular novelties include:
- The most reliable synthetic relational data on the market: In recent publications, we outline our patented methods for generating synthetic relational data 123. Aindo Synth 3.1 is the first version of the platform in which these methods are fully incorporated. Aindo Synth therefore outperforms the state-of-the-art, particularly for relational databases with many tables and connections.
- Improved data management: Aindo Synth 3.1 also includes new features for managing both tabular and relational datasets. This includes effortless storage of real and synthetic datasets in a wider range of formats. For relational datasets, management functions also include automated visualization of relationships between tables. Users can also manually remove, add and alter connections for improved accuracy. This allows them to refine the structure of a relational database to better reflect relationships between entities. It further allows for the elimination of redundant or irrelevant connections that could lead to incorrect or misleading results.
- Synthetic data with deterministic patterns: Aindo Synth 3.1 includes an innovative, effective method to accurately preserve both stochastic and deterministic patterns in synthetic datasets. Databases may involve causal dependencies between attributes. For example, in a medical dataset, diastolic blood pressure is necessarily lower than systolic blood pressure. While state-of-the-art synthetic data technologies are good at inferring probabilistic patterns, they generally cannot preserve such fixed necessities. In Aindo Synth 3.1, both statistical and deterministic patterns are reproduced with an uncanny degree of realism.
Italian Patent No. 102021000008552 ↩
Daniele Panfilo, (2022). Generating Privacy-Compliant, Utility-Preserving Synthetic Tabular and Relational Datasets Through Deep Learning. University of Trieste. ↩
Ciro A. Mami, Andrea Coser, Eric Medvet, Alexander T.P. Boudewijn, Marco Volpe, Michael Whitworth, Borut Svara, Gabriele Sgroi, Daniele Panfilo and Sebastiano Saccani (2022). Generating Realistic Synthetic Relational Data through Graph Variational Autoencoders. Proceedings of the Thirty-sixth Conference on Neural Information Processing Systems (NeurIPS). ↩