Avoiding Loss of Quality in Tiny Models - Neuton.ai Partner Session

Explore the challenges and solutions in creating compact machine learning models for edge devices in this tinyML EMEA 2021 Partner Session. Dive into Neuton.ai's approach to balancing model size and accuracy, evaluating model quality, and ensuring explainability in neural networks. Learn how to assess training data, interpret model decisions, and identify key parameters for building efficient tiny models. Discover techniques for monitoring model performance, detecting decay, and evaluating prediction credibility. Gain insights into customization, hardware considerations, and the future of tinyML implementation through this comprehensive presentation by Blair Newman, CTO of Neuton.ai.

Intro
Summary
Platform Overview
Prediction Tab
Model Data Relevance Indicator
Customization
Compact models
Sensors
MCUs
Floats
QA
Roadmap
Cloud dependency
Time to iteration
Big data
Time taken for iteration
Ideal edge hardware device
Ideal collaborators
Closing remarks
Did you try to implement the generated model
Top questions
Reconfigurable spiking neural network
Conclusion