pi may

Uploaded from authorPOINT
Views:
 
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Adaptive Model-Integrated Computing: 

Adaptive Model-Integrated Computing Akos Ledeczi

The MultiGraph Architecture: 

The MultiGraph Architecture

Introduction: 

Introduction Goal: to create software that continuously evaluates its own performance and adapts itself to meet performance goals Target: embedded signal processing / instrumentation systems that can be structurally adapted Method: models and model interpreters are embedded into the runtime system

Slide4: 

Adapted System Input Output Architecture Adapter

Slide5: 

Runtime System Evaluator Configurator System Models Input Output Architecture

Generative Modeling: 

Generative Modeling Need to model the solution space not just a single solution Parametric description of model structures Numerical parameters + generator algorithms Paradigm independent to support reuse Declarative + Generative Modeling

Slide7: 

Active Models Evaluator Runtime System Embedded Translator Embedded Models ASC Modeling Environment ASC Translator Parameters Design Time Run Time Prototype Architecture Input Output

System Validation: 

System Validation Validation in design time is infeasible Constrain the target system, not the models (e.g. resource constraints) Constraints prune the search space Iterative adaptation

Slide9: 

Active Models Evaluator Runtime System Embedded Interpreter Embedded Models Parameters System Validation Input Output Constraint Checker Constraints

Slide10: 

FFT Simple Adaptive Spectrum Analyzer Peak Detector Evaluator Input Reconfiguration Filter Bank

Adaptive Spectrum AnalyzerDemonstration: 

Adaptive Spectrum Analyzer Demonstration

Meta-Model of ASC paradigm: 

Meta-Model of ASC paradigm

Model of Spectrum Analyzer: 

Model of Spectrum Analyzer

Generative Modelof Adaptive Filterbank: 

Generative Model of Adaptive Filterbank

Analyzer Demonstration: 

Analyzer Demonstration Measured frequency (FFT frequency estimates in parentheses)

Signal Setup: 

Signal Setup

Conclusions: 

Conclusions Modeling: Declarative and Generative Models Active Models Embedded Models, Constraints and Translators Reconfiguration Evaluation System validation Real time issues Transients at reconfiguration