YoVDO

Ekya - Continuous Learning of Video Analytics Models on Edge Compute Servers

Offered By: USENIX via YouTube

Tags

USENIX Symposium on Networked Systems Design and Implementation (NSDI) Courses Continuous Learning Courses Edge Computing Courses Data Drift Courses

Course Description

Overview

Explore a conference talk on Ekya, a system for continuous learning of video analytics models on edge compute servers. Dive into the challenges of balancing inference and retraining tasks, addressing data drift, and optimizing resource allocation. Learn how Ekya outperforms baseline schedulers, achieving 29% higher accuracy gain and requiring 4x fewer GPU resources. Discover the intricacies of edge video analytics, the cost of continuous learning, and innovative scheduling decisions. Gain insights into Ekya's design, including its Thief Scheduler, and understand its performance in scaling with increasing video streams.

Syllabus

Intro
Video data is everywhere
Why video analytics at the edge?
Edge Video Analytics Setup
The cost of continuous learning
Resource demands of continuous learn
Summary thus far
Scheduling decisions to make
Working Example
Example - Fair Scheduler
Example - a smarter schedule
Key Takeaways
Ekya Design
Ekya Thief Scheduler Goal: Maximize mean inference accuracy across all jobs
Evaluation
Scaling with increasing video streams


Taught by

USENIX

Related Courses

How to Detect Silent Failures in ML Models
Data Science Dojo via YouTube
Dataset Management for Computer Vision - Important Component to Delivering Computer Vision Solutions
Open Data Science via YouTube
Testing ML Models in Production - Detecting Data and Concept Drift
Databricks via YouTube
Building and Maintaining High-Performance AI
Data Science Dojo via YouTube
Monitoring ML Models - Full Stack Deep Learning - Spring 2021
The Full Stack via YouTube