embARC Machine Learning Inference (MLI) Workshop

ARC Processor Summit - Silicon Valley

Thursday, September 19, 2019
Santa Clara Marriott
2700 Mission College Blvd, Santa Clara, CA 95054

Session 1: 11:30 a.m. – 1:30 p.m.
Lunch will be served from 1:30 p.m. – 2:15 p.m.
Session 2: 3:15 p.m. – 5:15 p.m.
Lunch will be served at 1:00 p.m.

Instructors: Jamie Campbell , Anatoly Savchenkov & Dmitry Zakharov

The embARC Machine Learning Inference (MLI) software library is optimized for low-power IoT applications that utilize convolutional neural networks (CNN) and recurrent neural networks (RNN). During this workshop, participants will get hands-on experience using the MLI library on the Synopsys ARC EM processor, by building an application which uses a CNN to recognize hand-written characters.

Instructors will provide step-by-step instructions and “check points” along the way so that everyone can stay on schedule.

Attendees will be able to keep the instructions and the software content will be open-sourced.

Attend this workshop to learn:
• How to implement efficient machine learning “at the edge” using Synopsys’ ARC EM processor
• How Synopsys’ optimized open-source Machine Learning Inference libraries target low-power ARC processors
• How to map neural network graphs defined in the TensorFlow framework to low-power ARC processors
• How to create a real-time character-recognition application that runs on ARC EM SDP hardware
• How to work with the ARC MetaWare Development Toolkit to build and debug applications running on the ARC EV processor

*Attendees should have C/C++ programming experience and some basic knowledge of Python.

In this hands-on workshop, you'll:
• Gain an introductory understanding of the embARC MLI Library and basic usage concepts
• Prepare a character-recognition CNN graph using TensorFlow
• Transform and quantize of the graph into an “MLI-ready” format
• Integrate the model into an embedded application suitable for execution on the ARC EM-based target
• Execute and test the application