EDP discovery from multi-source data via competitive optimization

The discovery of fundamental laws from experimental data has been a central goal of computational science for decades. In recent years, artificial intelligence-based approaches have made it possible to extract partial differential equations (PDEs) that describe complex physical systems. However, most of these methods operate on a single dataset, which limits their accuracy when observations are scarce or noisy. A significant advance in this field proposes a competitive optimization framework that integrates multiple data sources, each with different initial or boundary conditions, to robustly discover the underlying shared PDEs. This approach, named MCO-PDE, first trains independent neural models for each source and then employs a soft weighting mechanism that dynamically evaluates the credibility of each dataset, merging global consensus coefficients. Combined with genetic algorithms for structural search, the system is capable of identifying both the functional form and the parameters of the governing laws, even in two-dimensional and three-dimensional domains with irregular boundaries and heterogeneous coefficients. The results demonstrate that with just 50 observations per source, canonical equations are recovered with high fidelity, and it has been validated in real wave tank experiments. From a business and technological perspective, this type of advancement represents an opportunity to develop tailored applications that accelerate scientific research and engineering. At Q2BSTUDIO we offer artificial intelligence solutions for businesses that can integrate symbolic discovery techniques and heterogeneous data fusion, adapting to sectors such as energy, manufacturing, or biomechanics. Our AWS and Azure cloud services provide the scalable infrastructure needed to process large volumes of information from multiple sensors or simulations, while our AI agent capabilities allow automating the complete cycle of experimentation and model validation. Likewise, the implementation of custom software facilitates the personalization of these competitive optimization frameworks for the specific needs of each organization. The combination of business intelligence with tools such as Power BI makes it possible to visualize and monitor data quality and algorithm convergence, ensuring informed decisions. In a context where cybersecurity is critical when handling proprietary or sensitive data, our data protection services ensure the integrity of processes. This paradigm of automated scientific discovery not only accelerates the acquisition of knowledge, but also opens the door for companies and institutions to adopt AI strategies beyond prediction, delving into the causal understanding of phenomena. Multi-source data fusion, optimized through competition between sources, is an example of how advanced artificial intelligence can solve problems that previously required intensive human intervention. At Q2BSTUDIO we accompany our clients in the adoption of these technologies, from conceptualization to production deployment, with a practical and results-oriented approach.