2019 International Conference On Computer Aided Design

The Premier Conference Devoted to Technical Innovations in Electronic Design Automation

November 4-7, 2019The Westin Westminster Westminster, CO

MP Associates, Inc.
THURSDAY November 07, 8:00am - 12:00pm | Westminster I
EVENT TYPE: WORKSHOP
SESSION 2W
Hands On Workshop on Machine Learning, Deep Learning and Reinforcement Learning for EDA Developers

Speakers:
Manish Pandey - Synopsys, Inc.
Claudionor Coelho - Google, Inc.
Organizer:
Claudionor Coelho - Google Inc.
This workshop covers the basics of machine learning, deep learning and reinforcement learning with examples on how to use this information to build the next generation of EDA tools. Machine learning is gaining greater acceptance in EDA industry, replacing complicated heuristics and solving problems that were very hard to characterize without a probabilistic mindset, with applications in design, verification, physical implementation and debugging. This workshop shows how to solve common EDA problems using Naïve Bayes, regression and classification, going to convolutional and recurrent neural networks, and finally presenting reinforcement learning, in a completely hands on manner, so that the attendees will have first hand experience in solving EDA problems with ML/DL/RL formulations. We will also focus on how to debug these problems and understand what goes on when the results do not match what you expected. During the workshop, we have selected a number of small EDA projects that are going to be executed by the attendees. 1. Regression and classification of parameters in NP complete heuristics 2. Estimating delays of libraries using machine learning and deep learning techniques 3. Using reinforcement learning to solve EDA problems 4. Using recurrent networks to analyze temporal data in EDA Attendees will receive an executable notebook containing all the material, with examples both in Keras and PyTorch. At the end of the workshop, attendees will be able to use ML/DL/RL to solve the most challenging problems in EDA, being able to discuss, implement and architect solutions.