Model Aware Computing with Scientific Categories
Abstract
Major Goals: The foundational problem is that of fusing models expressible in disparate Domain Specification Language (DSL) to be able to draw meaningful inferences. The PI will bring Category Theory (and rich construction mechanisms) that would mediate/ act as a glue to combine disparate models to provide a uniform framework to build decision aide systems. The work will be carried out in the context of DARPA World Modelers program, and the DARPA ASKE (Automatic Scientific Knowledge Extraction) program, which will provide data to validate the proposed work. Accomplishments: The largest area of technical progress was the completion of 3 papers that were submitted (see below for details). As far as new technical capabilities, we have made a significant achievement in the implementation of double pushout rewriting for arbitrary C-Sets. This capability depends on a novel generalization of techniques used in the graph rewriting community to a much broader class of data structures. This algorithm shows why C-Sets are an ideal mathematical setting for reasoning about modeling frameworks. They are an incredibly general class of data structures (known to be isomorphic to relational databases), which still allow for the generic implementation of complex algorithms. DPO rewriting can be used to model many processes that manipulate data structures and thus implement many algorithms. For example as part of the ACT Adjoint School, we implemented the classical Ising model of magentic spin systems with DPO rewriting, where the states of the system are stored as C-Sets and the DPO rules implement the dynamics of the Ising model. These rules capture the fundamental process of the Ising model which is to flip individual spins. The work with CSet DPO rewriting has allowed us to implement model space exploration for Petri Nets (which are CSets) and we have explored how DPO rules allow us to capture the intuitive rules that scientists can use to compare model structures.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2021
- Accession Number
- AD1209143
Entities
People
- Margaret L. Loper
Organizations
- Georgia Tech