#Japan #AI #Shibuya #Velt #xCausal®

Velt's Innovative Approach to Tackling Data Scarcity with xCausal®

In the modern landscape of AI and statistical analysis, a pressing issue has emerged: the scarcity of specific data sets. Velt Co., Ltd., based in Shibuya, Tokyo, has recognized this challenge and is now addressing it head-on with the launch of their new feature, the "xCausal® Data Augmentation Function," which began offering on July 13th. This feature is a component of their causal AI platform, xCausal®, and allows for the generation of extensive datasets from minor seed data, tailored for AI development and statistical analysis.

Understanding the Technology Behind xCausal®

Utilizing Graphical Causal Models (GCMs) as a foundation, the data augmentation function captures the realities of target domains even with minimal seed data. This approach is particularly vital in fields such as healthcare research for rare diseases, where numerous variables may exist, yet cases remain scarce. The functionality extends beyond healthcare, also serving sectors like manufacturing, finance, infrastructure, and mobility, where data shortages are often critical.

The xCausal® feature provides a unique solution by integrating a fidelity verification mechanism, allowing users to quantitatively assess how faithfully the generated data reflects the original seed data.

A Radical Solution to Data Scarcity

The heart of the issue lies in the structural scarcity of data across various industries. Critical events—such as defects in products, fraudulent transactions, or equipment malfunctions—tend to occur infrequently, resulting in an uneven distribution of data. As a result, many predictive models fail to catch important cases due to this inherent bias. Meanwhile, generative AI methods that have become popular recently require vast amounts of training data, making them less effective in data-scarce contexts.

Leveraging Causal Models for Data Generation

Traditional data generation techniques like GANs and VAEs focus on mimicking observed correlations. However, they fall short of addressing underlying causal relationships, which can diminish their effectiveness when sufficient reference data is not available. In contrast, GCMs offer a systematic blueprint of how data emerges from causal linkages. For example, they can define processes where changes in certain conditions lead to variations in outcomes, paving the way for coherent data generation from limited seed data. This method not only models familiar conditions but can also produce plausible data for unobserved scenarios, enhancing its applicability in fields that necessitate justification and transparency like healthcare and finance.

Key Features of the xCausal® Data Augmentation

1. Causal Structure-Based Data Expansion
Unlike common methods that interpolate based on observed correlations, the xCausal® approach generates data grounded in causal structures, thus making it effective even from few seed dimensions. This enables the creation of counterfactual scenarios for rare event contingencies or stress testing.

2. Fidelity Verification Tools
The augmented data's fidelity to the seed data is quantitatively verified through two perspectives: the similarity of individual variable distributions and the preservation of correlations between variables. For more rigorous validation, predictive model functionality can be offered as a professional service, ensuring the augmented data's utility against real-world performance.

3. User-Friendly Platform
Built on the xCausal® platform, which prioritizes ease of use for business users, this functionality comes with robust tools for causal inference, robustness evaluation, and root cause analysis—all without needing coding expertise.

4. Professional Support Services
Velt offers comprehensive professional services ranging from building causal models to validating the utility of augmented data, ensuring organizations maximize the potential of their newly generated datasets.

Contributing to AI-Ready Data

The recent surge in demand for AI-Ready data, characterized by high accuracy, consistency, and completeness, presents another layer of significance for the xCausal® feature. With the Japanese government outlining its initiatives for a “trustworthy AI,” Velt's approach helps meet these needs by expanding small sets of quality seed data into larger datasets without sacrificing their integrity.

The xCausal® Data Augmentation Function stands poised to address the constraints that have traditionally limited AI's application due to data shortages, assisting various industries in creating productive cycles of data and AI.

Broad Application Opportunities

Velt is preparing to deploy this function across various fields, including:

• - Healthcare and Drug Development: Generating external control arms for clinical trials from a handful of patient data.
• - Financial Services: Simulating anti-money laundering scenarios to uncover previously unobserved fraud patterns.
• - Physical AI: Producing simulated accident data that adheres to physical laws from minimal observational data.

These case studies illustrate how xCausal® could revolutionize data utilization across sectors where data scarcity has been a critical challenge.

Conclusion

As Velt continues to evolve its mission—"We power your business with WHY: Redefining trust in the age of AI"—the xCausal® Data Augmentation Function represents a profound step forward in empowering businesses to harness data effectively, driving decision-making based on a deeper understanding of causal relationships. The technology promises to not only enhance current data analytics capabilities but also reshape how organizations interact with their data ecosystems in the future.

For more information, visit Velt's official website at https://veldt.jp.

About Velt Co., Ltd.
Founded in August 2012, Velt is at the forefront of leveraging causal AI technologies to solve societal challenges and accelerate good-will innovation.