Complex systems like those in biology, economics, or ecology are difficult to represent and properly understand. Complex interactions, emergent properties, and high dimensionality make it difficult to disentangle the governing rules and systemic properties when trying to model a given environment. We often depend on external tools to help us see particular patterns that are otherwise non-salient. These tools help us to better understand narrow relationships, aggregate and filter large sets of data, run simple analytics, and to simplify complicated networks and processes.
Despite the vast expanse of available data and computing power, current predictive modeling remain reductionist in their approach to modeling these complex systems. Typical heuristic solutions are effective for tasks like natural language processing, anomaly detection, or recognition/categorization; these techniques, however, must be applied to specific uses and require fundamental understanding about the parameters and desired outcomes. Microscale modeling techniques like Agent Based Modeling (ABM) are capable of representing systems of interconnected entities to elucidate emergent principals, but require defining of relational properties. Yupana attempts to combine the principles from these processing and modeling techniques into a cohesive and modular framework.
Yupana seeks to implement the precision of machine learning into the distributed format of ABM without setting or relying on an end-goal. To accommodate drastic changes in the environment that would otherwise render a simulation model obsolete, Yupana will need to consume asynchronous data updates from customizable data streams. The simulation framework of this project must be developed to adapt to many different systems of inquiry. This means allowing integration of an array of different processor types, as well as easy integration of new sources and modeling parameters.
Comments or questions? Join the discussion at BlockShop!