Resilient Edge-Cloud Autonomous Learning with Timely inferences

Project Overview and Journey

Project Advisor and Mentors: Professor Anand Sarwate, Professor Waheed Bajwa, Nitya Sathyavageeswaran, Vishakha Ramani, Yu Wu, Aliasghar Mohammadsalehi, Muhammad Zulqarnain

Team: Shreya Venugopaal, Haider Abdelrahman, Tanushree Mehta, Lakshya Gour, Yunhyuk Chang

Objective: The purpose of this project is to use ORBIT in order to design and run experiments which will analyze the training and prediction of various ML models across multiple edge devices. Additionally, students will develop a latency profiling framework for MEC-assisted machine learning using cutting edge techniques such as splitting models across multiple orbit nodes and networks, as well as early exit. They will then analyze these models for latency and accuracy tradeoff analysis, along with measuring network delays.

Table of Contents / Links to Presentations

• Week 1
• Week 2
• Week 3
• Mid-Sem

Weekly Progress

Week 1

Goals
• 1
• 2
• 3
• 4

Summary
• 1
• 2
• 3
• 4

---

Week 2

Goals
• 1
• 2
• 3
• 4

Summary
• 1
• 2
• 3
• 4

---

Week 3

Goals
• 1
• 2
• 3
• 4

Summary
• 1
• 2
• 3
• 4

---

Week 4

Goals
• 1
• 2
• 3
• 4

Summary
• 1
• 2
• 3
• 4

---

Goals

• Team introductions
• Kickoff meetings
• Refining and understanding research questions
• go over goals for the summer

For next week:

• Go through WINLAB orientation (TBD)
• Reading for next time:
• PPA Chapters 1 and 2
• Coding:
• install python (via Anaconda or something else)
• go through the interactive intro to python online or by downloading the notebook
• Review the basics of Python linked from the UT Austin site to learn basic syntax, flow control, etc.

*Week 2: Goals: Summary:

• Basics of pattern recognition and Machine Learning (PPA - Patterns, Predictions, Actions)
• set up an instance using pytorch on an Orbit node
• Created a node image with Pytorch
• Basics of Pytorch
• Created small Machine Learning models
• Loaded the Modified National Institute of Standards and Technology (MNIST) database onto the node

computed the second moment matrix

• Did PCA and SVM on MNIST

Next Steps:

• Create and train a “small” and “large” Neural Network
• Attempt to simulate the difference between their performances at inference

Links: Week 3: Goals: Summary:

• Created and Trained a ‘Small’ and ’Large’ Neural Networks
• Compared their performances on the CIFAR10 dataset
• Established connection between two nodes
• Communicated test data between nodes to compare accuracy and delay between our NN models
• Need to serialize by bytes instead of transferring as strings
• Had a discussion on various research papers related to our project

Next Steps:

• Send the data as bytes instead of strings
• Calculate the times for transfer & processing
• Read more papers related to stuff about early exit and edge cloud computing
• Divide data into “chunks” for faster and more efficient transmission
• Design experiments to test different architectures and implementations of Early Exiting and Split Computing
• Track and add Age of Information Metrics

Presentation links:

• Link - Lakshya
• Link - Shreya
• Link - Haider
• Link - James
• Link - Tanushree

Week 4: Goals Summary Next Steps