Inhalt

•  
   Acknowledgements
•  
   Abstract
•  
   Zusammenfassung
•  
   Table of Contents
•  
  Introduction
•  
   Challenges for CMOS technology in the Nano-era
•  
   Approximate computing
•  
   Approximate computing at different layers of system stack
•  
   Approximate arithmetic blocks
•  
   The focus of this thesis
•  
   Contributions of the thesis
•  
   Organization of the remainder
•  
  Background and Related Work
•  
   Automated synthesis of approximate accelerators
•  
   Approximation instance generation
•  
   Search space exploration
•  
   Quality evaluation
•  
   Error metrics
•  
  Related works
•  
   Search-based methods
•  
   Analytical methods
•  
  MCTS-based Framework for Approximate Accelerator Synthesis
•  
   Chapter overview
•  
   Proposed MCTS-based framework for approximate accelerator synthesis
•  
   Overview of the framework
•  
   MCTS and its adaptation to the accelerator circuit synthesis
•  
   Selection policy
•  
   Search space expansion policy
•  
   Selection of approximate transformations
•  
   Reward function
•  
  Experimental setup and results
•  
   Setup of experiments
•  
   Results and discussion
•  
  Towards a modular and flexible framework CIRCA
•  
   Motivation behind development of CIRCA
•  
  Concept and architecture of CIRCA
•  
   Input stage
•  
   The QUAES stage
•  
   The output stage
•  
  Search space exploration methods in CIRCA
•  
   Hill climbing (HC)
•  
   Simulated annealing (SA)
•  
   Monte Carlo tree search (MCTS)
•  
   Bounding the search space
•  
   Experimental setup and results
•  
   Setup of experiments
•  
   Results
•  
   Further Discussion
•  
   Chapter conclusion
•  
  Hybrid Methodology for Synthesis of Approximate Accelerators
•  
   Chapter overview
•  
   Motivational example
•  
   Proposed hybrid methodology
•  
  Analytical bit-width estimation through EVT
•  
   Extreme value theory
•  
   Analytical bit-width estimation via EVT
•  
  Parallel search-based optimization
•  
   Parallel MCTS
•  
  Experimental results
•  
   Setup of experiments
•  
   Results and discussion
•  
   Chapter conclusion
•  
  Machine Learning-based MCTS
•  
   Chapter overview
•  
   Motivation and background
•  
   DNN enabled MCTS-based AxAC synthesis framework
•  
   Deep neural network based error estimation regressor models
•  
   Design space exploration via MCTS
•  
  Experimental results
•  
   Setup of experiments
•  
   Results and discussion
•  
   Chapter conclusion
•  
  Conclusion
•  
   Summary
•  
   Future directions
•  
   List of Tables
•  
   List of Figures
•  
   Author's Publications
•  
   Bibliography