On a federated architecture for utilizing domain-specific descriptive models. (Poster)
원문 초록
In a Systems Engineering setting, various models are produced using a variety of methods and tools. Focusing on a type of models -- called descriptive models -- which we shall describe, we argue that, while the clarity and precision of models are essential for their exchange and reuse, the way in which the data of these models are defined and the information conveyed is also essential for their (re) utilization -- e.g., for analysis or synthesis purposes. Category Theory has made it possible to link seemingly separate fields or domains, so that anything that can be rewritten in this framework benefits from a level of abstraction and a relational viewpoint -- essential to address complexity. We therefore take advantage of this framework to define a federated architecture for projecting and conveying these models without sacrificing clarity and precision. A federated architecture has two important advantages. On the one hand, it unifies these models from their structure; on the other hand, it allows a specific usage (instantiation or interpretation) of this structure within a business domain. We define the structure of these models as a symmetric multicategory. In particular, we rely on matrices over a semiring to define morphisms and their composition. The choice of matrices is intended to facilitate the application in practice.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Describe Anything Model for Visual Question Answering on Text-rich Images
Assay2Mol: large language model-based drug design using BioAssay context
Beyond Accuracy: Metrics that Uncover What Makes a 'Good' Visual Descriptor
Dataset Distillation via Vision-Language Category Prototype
DRAMA-X: A Fine-grained Intent Prediction and Risk Reasoning Benchmark For Driving
InstructTTSEval: Benchmarking Complex Natural-Language Instruction Following in Text-to-Speech Systems
토론
이 논문과 연결된 포럼 글이 아직 없습니다.