Presentation
Adding machine learning to the mix of EDA optimization algorithms
TimeMonday, December 6th10:30am - 12:00pm PST
Location3016
Event Type
Tutorial
Presented In-Person
DescriptionAdvances in computing power and the availability of big data have ignited innovations in EDA. These advances rely on efficient algorithms that can produce optimal results in a short period of time. Many of the existing algorithms take advantage of mathematical optimization techniques to improve their solution quality. These techniques can find optimal solutions and require some level of insight into the nature of the problem by the designer. However, mathematical optimization relies heavily on the developed models and is often not robust concerning the variations in the presence of large data sets. That is why machine learning and deep learning techniques have gained popularity. Machine learning techniques can solve large-scale problems efficiently once they are trained. These methods do not use traditional modelling approaches and are often not sufficiently understood.
The main purpose of this tutorial is to show how optimization and machine learning can be used in a virtuous cycle to use the data to develop better models and use the models to generate more data to improve the accuracy and robustness of our models and develop with more innovative methods to solve the very challenging problems we face today. We would also like to address the false dichotomy that it is either optimization or machine learning and show how by combining these two methods we can improve our results. We will demonstrate the application of this approach to several EDA problems including estimation models during placement, smart solution space exploration during synthesis and physical design, improving convergence of iterative algorithms by running an ML model on the solution from the previous iteration
In this interactive tutorial, we will discuss the algorithm-driven nature of the optimization techniques and compare that to the data-driven nature of the machine learning techniques. Then, we will discuss how optimization and ML can be made into a virtuous cycle to solve the problems of scaling both in numbers and transistor sizes. We will show some examples of EDA problems and how they have been solved using optimization, machine learning, or both.
The main purpose of this tutorial is to show how optimization and machine learning can be used in a virtuous cycle to use the data to develop better models and use the models to generate more data to improve the accuracy and robustness of our models and develop with more innovative methods to solve the very challenging problems we face today. We would also like to address the false dichotomy that it is either optimization or machine learning and show how by combining these two methods we can improve our results. We will demonstrate the application of this approach to several EDA problems including estimation models during placement, smart solution space exploration during synthesis and physical design, improving convergence of iterative algorithms by running an ML model on the solution from the previous iteration
In this interactive tutorial, we will discuss the algorithm-driven nature of the optimization techniques and compare that to the data-driven nature of the machine learning techniques. Then, we will discuss how optimization and ML can be made into a virtuous cycle to solve the problems of scaling both in numbers and transistor sizes. We will show some examples of EDA problems and how they have been solved using optimization, machine learning, or both.